‫ﻣﻮﻟﻮﺩ ﺣﺸﻤﺎﻥ‬

‫اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ وﺗﻘﻨﻴﺎت‬

‫اﻟﺘﻨﺒﺆ اﻟﻘﺼﲑ اﳌﺪى‬
‫‪@ @ ÒÜÌåflÎ@Úz‘‰fl@Úr¤bq@Ú»jü‬‬
2
‫ﺑﺳم ﷲ اﻟرﺣﻣن اﻟرﺣﯾم‬
 ‫‪@ @êäË–¤a‬‬
‫‪11‬‬ ‫‪@ÚflÜ‘fl‬‬

‫‪13‬‬ ‫‪@ÚÌ˚j‰n¤a@xáb‡‰¤a@Z@fiβa@›ñ–¤a‬‬
‫‪13‬‬ ‫‪ -1‬ﺗﻌﺎﺭﻳﻒ‬
‫‪16‬‬ ‫‪ -2‬ﺃﻧﻮﺍﻉ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺘﻨﺒﺆﻳﺔ‬
‫‪17‬‬ ‫‪ - 1-2‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ‬
‫‪19‬‬ ‫‪ - 2-2‬ﳕﺎﺫﺝ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‬
‫@ @‬
‫‪@ @ lbđ‘nç ¸a@xáb∏@Zfiβa@lbj¤a‬‬

‫‪25‬‬ ‫‪@bË–í◊@÷äü@Î@ÚÓ‰flå¤a@@Ú‹é‹é¤a@pbj◊äfl@ZÔ„br¤a@›ñ–¤a‬‬
‫‪25‬‬
‫‪ -1‬ﺍﳌﺮﻛﺒﺎﺕ ﺍﳉﻮﻫﺮﻳﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫‪29‬‬ ‫‪ -2‬ﻛﺸﻒ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫‪29‬‬ ‫‪ -1-2‬ﺍﻟﺘﺤﻠﻴﻞ ﺍﻟﺒﻴﺎﱐ ﻟﻠﻤﻌﻄﻴﺎﺕ‬
‫‪30‬‬ ‫‪ -2-2‬ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﻹﺣﺼﺎﺋﻴﺔ‬
‫‪30‬‬ ‫‪ -1-2-2‬ﻛﺸﻒ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‬
‫‪46‬‬ ‫‪ -2-2-2‬ﻛﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ‬
‫@ @‬
‫‪61‬‬ ‫‪ÚđÓéj¤a@lbđ‘nç ¸a@xáb‡‰i@˚j‰n¤a@Zs¤br¤a@›ñ–¤a‬‬
‫‪61‬‬ ‫‪ -1‬ﺍﻟﺘﻨﺒﺆ ﺑﻨﻤﺎﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﻃﺮﻕ ﺗﻘﻴ‪‬ﻴﻤﻬﺎ‬

‫‪5‬‬
‫‪84‬‬ ‫‪ -2‬ﺍﻟﺘﻨﺒﺆ ﻭﺍﻟﺘﻤﻬﻴﺪ ﺑﺎﻟﻨﻤﺎﺫﺝ ﺍﳌﻜﻴﻔﺔ‬
‫‪84‬‬ ‫ا‪ -‬ﺍﻟﺘﻨﺒﺆ‬
‫‪88‬‬
‫ب‪ -‬ﺍﻟﺘﻤﻬﻴﺪ‬

‫‪103‬‬ ‫‪@ÚÓ‹ñ–¤a@Új◊äΩa@Z…iaä¤a@›ñ–¤a‬‬
‫‪104‬‬ ‫‪ -1‬ﺍﻹﺯﺍﻟﺔ ﻭﻃﺮﻗﻬﺎ‬
‫‪121‬‬
‫‪ -2‬ﺍﻟﻨﻤﺬﺟﺔ‬

‫‪ÚÓˆaÏí»¤a@ÚÓ‰flå¤a@Ú‹é‹é¤a@›Ó‹§@ZÔ„br¤a@lbj¤a‬‬

‫‪135‬‬ ‫‪@ @@Box-Jenkins @å‰Óÿ„bu M@ è‬‬ ‫‪◊Ïi@Ú‘Ìäü@Zè flb®a@›ñ–¤a‬‬

‫‪136‬‬ ‫‪ -1‬ﺧﺼﺎﺋﺺ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫‪146‬‬ ‫‪ -2‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﳋﻄﻴﺔ ﻟﻠﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‬
‫‪147‬‬ ‫‪ -1-2‬ﻣﺮﺣﻠﺔ ﲢﺪﻳﺪ ﺍﻟﻨﻤﻮﺫﺝ‬
‫‪147‬‬ ‫‪ -1-1-2‬ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‬
‫‪155‬‬ ‫‪ -2-1-2‬ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ‬
‫‪166‬‬ ‫‪ -3-1-2‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ‬
‫‪170‬‬ ‫‪ -4-1-2‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ ﺍﳌﺮﻛﺒﺔ‬
‫‪177‬‬ ‫‪ -2-2‬ﻣﺮﺣﻠﺔ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ‬
‫‪177‬‬ ‫‪ -1-2-2‬ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﳕﻮﺫﺝ ﺍﳓﺪﺍﺭ ﺫﺍﰐ‬
‫‪181‬‬ ‫‪ -2-2-2‬ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻭﺍﳌﺨﺘﻠﻄﺔ‪.‬‬

‫‪6‬‬
‫‪194‬‬ ‫‪ -3‬ﺗﺸﺨﻴﺺ ﺍﻟﻨﻤﻮﺫﺝ‬
‫‪203‬‬ ‫‪ -4‬ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﺍﻟﺸﻌﺎﻋﻴﺔ ﻭ ﺍﻹﺳﺘﻘﺮﺍﺭﻳﺔ‬

‫‪219‬‬ ‫‪@ @@Ú”Ü ¤a@êbÓ”Î@˚j‰n¤a@ZêÖbé¤a@›ñ–¤a‬‬

‫‪219‬‬ ‫‪ -1‬ﺍﻟﺘﻨﺒﺆ‬
‫‪224‬‬ ‫‪ -2‬ﻗﻴﺎﺱ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ‬
‫@ @‬
‫‪239‬‬ ‫‪@ @ÚÓˆbñyĂa@fiÎaܶa‬‬
‫‪247‬‬ ‫‪@ @…uaäΩa‬‬

‫‪7‬‬
8
 ‫@ @‬
‫ﻣﻘﺪﻣﺔ ﺍﳌﺆﻟﻒ‬

‫ﻳﺘﻨﺎﻭﻝ ﻫﺬﺍ ﺍﻟﻜﺘﺎﺏ ﺃﺳﺎﺳﻴﺎﺕ ﻋﻠﻢ ﺍﻻﻗﺘﺼﺎﺩ ﺍﻟﺘﻄﺒﻴﻘﻲ ﳑﺜﻼﹰ ﰲ ﻋﻠﻢ ﺍﻟﻘﻴﺎﺱ‬
‫ﺍﻻﻗﺘﺼﺎﺩﻱ ﺑﺸﻜﻞ ﻋﺎﻡ ﻭ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﻭ ﺍﻟﺘﻨﺒﺆ ﺑﺸﻜﻞ ﺧﺎﺹ‪.‬‬
‫ﻫﺬﺍ ﺍﻟﻜﺘﺎﺏ ﻣﻮﺟﻪ ﻟﻄﻠﺒﺔ ﻣﻌﺎﻫﺪ ﺍﻻﻗﺘﺼﺎﺩ ﻭﺍﻹﺣﺼﺎﺀ‪ ،‬ﻃﻠﺒﺔ ﺍﻟﻘﻴﺎﺱ‬
‫ﺍﻻﻗﺘﺼﺎﺩﻱ‪ ،‬ﺍﻻﻗﺘﺼﺎﺩ ﺍﻟﻜﻤﻲ‪ ،‬ﺍﻻﻗﺘﺼﺎﺩ ﺍﳌﺎﱄ ﻭﻛﺬﺍ ﻟﻄﻠﺒﺔ ﺍﻟﺪﺭﺍﺳﺎﺕ ﺍﻟﻌﻠﻴﺎ‪ .‬ﻟﺬﺍ‬
‫ﻓﺈﻥ ﻫﺬﺍ ﺍﻟﻌﻤﻞ ـ ﻓﻀﻼ ﻋﻦ ﺍﻟﻄﻠﺒﺔ ـ ﻓﻬﻮ ﻣﻮﺟﻪ ﻛﺬﻟﻚ ﺇﱃ ﺍﳌﺴﺆﻭﻟﲔ ﻭﺍﳌﺴﲑﻳﻦ‬
‫ﺳﻮﺍﺀً ﰲ ﺍﳌﺆﺳﺴﺎﺕ ﺍﻟﺼﻐﲑﺓ ﺃﻭ ﺍﻟﻜﺒﲑﺓ ﺑﻐﻴﺔ ﺗﺴﻬﻴﻞ ﺁﻟﻴﺎﺕ ﺍﻟﺘﺴﻴ‪‬ﲑ ﻭﺑﺸﻜﻞ ﺭﺋﻴﺴﻲ‬
‫ﻋﻤﻠﻴﺔ ﺍﲣﺎﺫ ﺍﻟﻘﺮﺍﺭ ﰲ ﻣﻴﺪﺍﻥ ﺍﳌﺒﻴﻌﺎﺕ ﻣﺜﻼ‪ ،‬ﺍﻟﺘﻜﺎﻟﻴﻒ‪ ،‬ﺍﻹﻧﺘﺎﺝ ﻭﺍﻟﺘﺸﻐﻴﻞ ‪...‬ﺍﱁ‪.‬‬
‫ﰎ ﺗﻘﺴﻴﻢ ﻫﺬﺍ ﺍﳌﻮﺿﻮﻉ ﺇﱃ ﺃﺑﻮﺍﺏ ﻣﺘﺒﺎﻳﻨﺔ‪ ،‬ﻭﺟﻪ ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ ﻟﻠﻤﺴﲑﻳﻦ ﻏﲑ‬
‫ﺍﳌﺨﺘﺼﲔ ﰲ ﻣﻴﺪﺍﻥ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ ﺧﺼﻮﺻﺎﹰ ﺃﻭ ﺍﻟﺘﻘﻨﻴﺎﺕ ﺍﻟﻜﻤﻴﺔ ﻋﻤﻮﻣﺎﹰ‪ ،‬ﻛﻮﻥ‬
‫ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﻣﻦ ﺍﻟﺘﻘﻨﻴﺎﺕ ﻭﺍﻟﻨﻤﺎﺫﺝ ﻻ ﲢﺘﺎﺝ ﺇﱃ ﺇﳌﺎﻡ ﻭﺍﺳﻊ ﲜﻮﺍﻧﺐ ﺍﻟﻨﻈﺮﻳﺔ‬
‫ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﳌﺘﻌﻠﻘﺔ ﲟﻮﺿﻮﻉ ﺍﻟﺪﺭﺍﺳﺔ ﻭ ﻻ ﺇﱃ ﺩﺭﺍﻳﺔ ﻛﺒﲑﺓ ﲟﻮﺿﻮﻉ ﻭﻋﻤﻠﻴﺎﺕ ﺍﻟﺘﻨﺒﺆ‬
‫ﻭﺫﻟﻚ ﻟﺒﺴﺎﻃﺔ ﺗﺮﻛﻴﺒﻬﺎ‪ ،‬ﻭﻣﻌﻘﻮﻟﻴﺔ ﺣﻠﻬﺎ ﻭﺳﻬﻮﻟﺔ ﺗﻔﺴﲑ ﻧﺘﺎﺋﺠﻬﺎ‪ ،‬ﻛﻤﺎ ﺃ‪‬ﺎ ﻻ ﲢﺘﺎﺝ‬
‫ﺇﱃ ﺇﻣﻜﺎﻧﻴﺎﺕ ﻛﺒﲑﺓ ﻣﺜﻞ ﺃﺟﻬﺰﺓ ﺍﻟﻜﻤﺒﻴﻮﺗﺮ ﺍﳌﺘﻄﻮﺭﺓ ﳌﺨﺘﻠﻒ ﺍﻟﻌﻤﻠﻴﺎﺕ ﺍﳊﺴﺎﺑﻴﺔ‪.‬‬
‫ﺑﻴﻨﻤﺎ ﺗﻘﻨﻴﺎﺕ ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ ﺗﻜﻮﻥ ﰲ ﺍﻟﻐﺎﻟﺐ ﻣﻮﺟﻬﺔ ﻟﻠﻤﺨﺘﺼﲔ ﰲ ﻫﺬﺍ‬
‫ﺍ‪‬ﺎﻝ‪ ،‬ﻭﻫﻲ ﺗﻌﺘﻤﺪ ﰲ ﺍﻷﺳﺎﺱ ﻋﻠﻰ ﺩﺭﺍﺳﺔ ﺍﻟﺴﻠﻮﻙ ﺍﳌﺎﺿﻲ ﻟﻠﻤﺘﻐﲑ ﺍﳌﺮﺍﺩ ﺍﻟﺘﻨﺒﺆ ﺑﻪ‬
‫ﻭﻓﻖ ﺃﺩﻭﺍﺕ ﲢﻠﻴﻞ ﻭ ﺃﺳﺲ ﺇﺣﺼﺎﺋﻴﺔ ﻗﻮﻳﺔ‪ ،‬ﰒ ﻭﻋﻠﻰ ﺃﺳﺎﺳﻬﺎ ﺍﻟﻘﻴﺎﻡ ﺑﻌﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﰲ‬
‫ﺍﳌﺴﺘﻘﺒﻞ‪ .‬ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﻣﻦ ﺍﻟﺘﻘﻨﻴﺎﺕ ﲢﺘﺎﺝ ﺇﱃ ﺇﻣﻜﺎﻧﻴﺎﺕ ﻻ ﺑﺄﺱ ‪‬ﺎ ﻟﻠﻘﻴﺎﻡ ‪‬ﺬﺍ‬
‫ﺍﻟﺪﻭﺭ ﺍﻟﻌﻤﻠﻲ ﻭﺍﻟﻔﻌﺎﻝ‪.‬‬

‫‪9‬‬
‫ﰲ ﺍﳋﺘﺎﻡ ﻭﺿﻤﻦ ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ ﺩﺍﺋﻤﺎ ﻭﰲ ﺍﻟﻔﺼﻞ ﺍﻟﺴﺎﺩﺱ ﻭﺍﻷﺧﲑ‪ ،‬ﻧﻠﻘﻲ‬
‫ﻧﻈﺮﺓ ﻋﻠﻰ ﻛﻴﻔﻴﺔ ﺍﻟﺘﻨﺒﺆ ﻭﻓﻖ ﳕﺎﺫﺝ ـ ﺑﻮﻛﺲ ﺟﻨﻜﻴﻨﺰ )‪ (BOX-JENKINS‬ـ ﻛﻤﺎ‬
‫ﻧﺴﺘﻌﺮﺽ ﺍﻟﻮﺳﺎﺋﻞ ﺍﻹﺿﺎﻓﻴﺔ ﻟﻘﻴﺎﺱ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ‪.‬ﻛﻤﺎ ﺃﺩﺭﺟﻨﺎ ﺿﻤﻦ ﻫﺬﺍ ﺍﻟﻔﺼﻞ‬
‫ﻣﻘﺪﻣﺔ ﺣﻮﻝ ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﺍﻟﺸﻌﺎﻋﻴﺔ ﺇﺿﺎﻓﺔ ﺇﱃ ﻣﺸﻜﻞ ﺍﻟﺘﻘﺪﻳﺮ ﺍﳌﺰﻳﻒ ﻭﻣﻨﻪ‬
‫ﺍﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‪ .‬ﻛﻤﺎ ﺍﺳﺘﻌﺮﺿﻨﺎ ﺃﻳﻀﺎ ﺑﻌﺾ ﺍﺧﺘﺒﺎﺭﺍﺕ ﺍﻻﺳﺘﻘﺮﺍﺭﻳﺔ‬
‫ﻭﺃﳘﻴﺘﻬﺎ ﰲ ﻣﻴﺪﺍﻥ ﺍﻻﺳﺘﻘﺮﺍﺭﻳﺔ ﻭﺍﻟﺘﻜﺎﻣﻞ ﺍﳌﺸﺘﺮﻙ‪ .‬ﻟﻘﺪ ﺣﺎﻭﻟﻨﺎ ﰲ ﻫﺬﺍ ﺍ‪‬ﺎﻝ‬
‫ﺍﻟﺘﺮﻛﻴﺰ ﻋﻠﻰ ﺍﳉﺎﻧﺐ ﺍﻟﺘﻄﺒﻴﻘﻲ ﻟﻠﻤﻮﺿﻮﻉ ﻭ ﺟﻌﻠﻪ ﰲ ﻣﺘﻨﺎﻭﻝ ﺍﻟﻄﻠﺒﺔ‪.‬‬
‫ﰲ ﺍﻷﺧﲑ ﺃﺟﺪ ﻧﻔﺴﻲ ﻣﺪﻳﻨﺎﹰ ﻟﻸﺳﺘﺎﺫ ﺍﳏﻤﺪ ﻋﻨﺎﻥ ﻋﻠﻰ ﻭﻗﻮﻓﻪ ﺍﳌﺴﺘﻤﺮ ﻋﻠﻰ‬
‫ﺗﻨﻘﻴﺢ ﺍﻟﻨﺴﺨﺔ ﺍﻷﻭﱃ ﻣﻦ ﺍﻟﻜﺘﺎﺏ ﻛﻤﺎ ﻻ ﺃﻧﺴﻰ ﺍ‪‬ﻬﻮﺩﺍﺕ ﺍﻟﻜﺒﲑﺓ ﺍﻟﱵ ﻗﺪﻣﺘﻬﺎ‬
‫ﺍﻵﻧﺴﺔ ﺃﻣﻴﻨﺔ ﺗﻴﻔﻮﺭﺓ ﰲ ﳎﺎﻝ ﺍﻟﺘﻨﻈﻴﻢ‪ ،‬ﺍﻟﻄﺒﺎﻋﺔ ﻭﺍﻹﺧﺮﺍﺝ ﳍﺬﻩ ﺍﻟﻨﺴﺨﺔ‪.‬‬

‫ﻣﻮﻟﻮﺩ ﺣﺸﻤﺎﻥ‬
‫ﻛﻠﻴﺔ ﺍﻟﻌﻠﻮﻡ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻭﻋﻠﻮﻡ ﺍﻟﺘﺴﻴﲑ ﺍﳉﺰﺍﺋﺮ‬
‫ﻣﺎﺭﺱ ‪2010‬‬

‫‪10‬‬
 ‫ﻣﻘﺪﻣﺔ‬

‫ﺇﺿﺎﻓﺔ ﺇﱃ ﺍﻟﻀﺮﻭﺭﺓ ﺍﻟﻌﻠﻤﻴﺔ‪ ،‬ﻭﺍﻟﻐﻴﺎﺏ ﺍﳌﻠﺤﻮﻅ ﳍﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﳌﺮﺍﺟﻊ ﻋﻦ‬

‫ﻣﻜﺘﺒﺎﺗﻨﺎ ﺑﺎﻟﻠﻐﺔ ﺍﻟﻌﺮﺑﻴﺔ‪ ،‬ﺃﺭﺩﻧﺎ ﲡﻤﻴﻊ ﻋﻨﺎﺻﺮ ﻫﺬﻩ ﺍﳌﺎﺩﺓ ﺑﺸﻜﻞﹴ ﻣﺒﺴﻂ‪ ،‬ﰲ ﻛﺘﺎﺏ‬
‫ﻳﺴﺘﻌﲔ ﺑﻪ ﺍﻟﻄﺎﻟﺐ ﰲ ﳐﺘﻠﻒ ﺍﳌﺴﺘﻮﻳﺎﺕ‪ ،‬ﺧﺎﺻﺔ ﺃﻧﻪ ﻳﻀﻢ‪ ‬ﲡﺮﺑﺔ ﺃﻛﺒـﺮ ﻣﻦ ﲦﺎﻧﻴﺔ‬
‫ﻋﺸﺮ ﺳﻨﺔ‪ ،‬ﻣﻦ ﺧﻼﻝ ﺗﺪﺭﻳﺴﻨﺎ ﳍﺬﻩ ﺍﳌﺎﺩﺓ ﺑﺎﻟﺬﺍﺕ ﰲ ﻣﻌﻬﺪ ﺍﻟﻌﻠﻮﻡ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ‬
‫ﺳﺎﺑﻘﺎﹰ‪ -‬ﺟﺎﻣﻌﺔ ﺍﳉﺰﺍﺋﺮ ﺳﺎﺑﻘﺎﹰ‪ -‬ﻟﻄﻠﺒﺔ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ ﻭﺍﻹﺣﺼﺎﺀ‪ ،‬ﻭﻃﻠﺒﺔ‬
‫ﺍﻻﻗﺘﺼﺎﺩ ﺍﳌﺎﱄ ﻭﻛﺬﺍ ﻟﻄﻠﺒﺔ ﺍﻟﺪﺭﺍﺳﺎﺕ ﺍﻟﻌﻠﻴﺎ‪.‬‬
‫ﺇﻥ ﻣﻮﺿﻮﻉ ﺍﻟﺘﻨﺒﺆ ﰲ ﺍﳌﻴﺪﺍﻥ ﺍﻻﻗﺘﺼﺎﺩﻱ ﺃﺧﺬ ﻭﻣﺎﺯﺍﻝ ﻳﺄﺧﺬ ﻗﺴﻄﺎﹰ ﻭﺍﻓﺮﹰﺍ‬
‫ﻣﻦ ﺍﻟﺪﺭﺍﺳﺔ ﻭﺍﻻﻫﺘﻤﺎﻡ‪ ،‬ﻧﻈﺮﺍﹰ ﻟﺘﻌـﻘﱡـﺪ ﺍﳊﻴﺎﺓ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﰲ ﻫﺬﺍ ﺍﻟﻌﺼﺮ‬
‫ﺑﺎﻟﺬﺍﺕ‪ ،‬ﻭﻛﺬﺍ ﺻﻌﻮﺑﺔ ﺇﺩﺍﺭﺓ ﺍﳌﺆﺳﺴﺎﺕ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﻟﻌﻤﻼﻗﺔ ﺇﺩﺍﺭﻳﺎﹰ ﺑﻀﺨﺎﻣﺔ‬
‫ﺣﺠﻢ ﻋﻤﺎﳍﺎ ﻭﺍﻗﺘﺼﺎﺩﻳﺎﹰ ﺑﺘﻨﻮﻉ ﻭﻛﱪ ﺣﺠﻢ ﺗﺸﻜﻴﻠﺔ ﻣﻨﺘﺠﺎ‪‬ﺎ ﺍﳊﺪﻳﺜﺔ‪.‬‬
‫ﻟﺬﺍ ﻓﺈﻥ ﻫﺬﺍ ﺍﻟﻌﻤﻞ ‪ -‬ﻓﻀﻼﹰ ﻋﻦ ﺍﻟﻄﻠﺒﺔ ‪ -‬ﻣﻮﺟﻪ ﻛﺬﻟﻚ ﺇﱃ ﺍﳌﺴﺌﻮﻟﲔ‬
‫ﻭﺍﳌﺴﲑﻳﻦ ﺳﻮﺍﺀً ﰲ ﺍﳌﺆﺳﺴﺎﺕ ﺍﻟﺼﻐﲑﺓ ﺃﻭ ﺍﻟﻜﺒﲑﺓ‪ ،‬ﺑﻐﻴﺔ ﺗﺴﻬﻴﻞ ﺁﻟﻴﺎﺕ ﺍﻟﺘﺴﻴ‪‬ﲑ‬
‫ﻭﺑﺸﻜﻞ ﺭﺋﻴﺴﻲ ﻋﻤﻠﻴﺔ ﺍﲣﺎﺫ ﺍﻟﻘﺮﺍﺭ ﰲ ﻣﻴﺪﺍﻥ ﺍﳌﺒﻴﻌﺎﺕ ﻣﺜﻼﹰ‪ ،‬ﺍﻟﺘﻜﺎﻟﻴﻒ‪ ،‬ﺍﻹﻧﺘﺎﺝ‬
‫ﻭﺍﻟﺘﺸﻐﻴﻞ‪...‬ﺍﱁ‪ ،‬ﻫﺬﺍ ﻣﻦ ﺟﻬﺔ ﻭﻣﻦ ﺟﻬﺔ ﺛﺎﻧﻴﺔ‪ ،‬ﲢﺴﲔ ﺍﻟﻨﺘﺎﺋﺞ ﺍﳌﺮﺗﻘﺒﺔ ﻟﻠﻘﺮﺍﺭﺍﺕ‬
‫ﺍﳌﺘ‪‬ﺨﺬﺓ ﻭﲢﺴﲔ ﻣﺴﺘﻮﻯ ﺍﻷﺩﺍﺀ ﰲ ﻫﺬﻩ ﺍﳌﺆﺳﺴﺎﺕ ﻭﺫﻟﻚ ﻷﻥ ﺍﲣﺎﺫ ﺍﻟﻘﺮﺍﺭ‬
‫ﺍﻋﺘﻤﺎﺩﺍﹰ ﻋﻠﻰ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﺑﻌﺪ ﺍﻻﺧﺘﻴﺎﺭ ﺍﳌﻮﻓﻖ ﳍﺎ‪ ،‬ﳜﻔﻒ ﺑﺸﻜﻞ ﻛﺒﲑ ﺍﳋﺴﺎﺭﺓ‬
‫ﺍﶈﺘﻤﻠﺔ ﺃﻭ ﺍﻟﺘﻘﻠﻴﻞ ﻣﻦ ﺣﺠﻢ ﺍﻟﻔﺮﺹ ﺍﻟﻀﺎﺋﻌﺔ‪.‬‬
‫ﺇﻻ ﺃﻧﻪ ﳝﻜﻦ ﺍﻹﺷﺎﺭﺓ ﺇﱃ ﺃﻥ ﻋﻤﻠﻴـﺔ ﺍﻟﺘﻨﺒـﺆ ﻭﺍﲣﺎﺫ ﺍﻟﻘﺮﺍﺭ ﰲ ﺍﳌﺆﺳﺴﺎﺕ‬
‫ﺍﻟﺼﻐﲑﺓ‪ ،‬ﻳﻘﻊ ﻋﻠﻰ ﻋﺎﺗﻖ ﺍﳌﺴﺌﻮﻝ ﺍﻷﻭﻝ ﻓﻴﻬﺎ‪ ،‬ﺑﻴﻨﻤﺎ ﰲ ﺍﳌﺆﺳﺴﺎﺕ ﺍﳌﺘﻮﺳﻄﺔ‬
‫ﻭﺍﻟﻜﺒﲑﺓ ﻻ ﳝﻜﻦ ﻟﻠﻤﺴﲑ ﺃﻥ ﻳﺘﺤﻤﻞ ﻟﻮﺣﺪﻩ ﻫﺬﺍ ﺍﻟﻌﺐﺀ ﺍﻟﻜﺒﲑ‪ ،‬ﻭﻟﺬﺍ ﻭﺟﺐ‬

‫‪11‬‬
‫ﺗﻜﻮﻳﻦ ﳐﺘﺼﲔ ﰲ ﻣﻴﺪﺍﻥ ﺍﻟﺘﻨﺒﺆ ﰒ ﺇﳚﺎﺩ ﻗﺴﻢ ﺃﻭ ﻫﻴﺌﺔ ﺧﺎﺻﺔ ﻭﻣﺘﺨﺼﺼﺔ ﳍﺬﺍ‬
‫ﺍﻟﻌﻤﻞ ﺍﻹﺳﺘﺮﺍﺗﻴﺠﻲ ﺍﳍﺎﻡ ﺳﻮﺍﺀً ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﺆﺳﺴﺔ ﺃﻭ ﺍﻟﺪﻭﻟﺔ‪.‬‬
‫ﳍﺬﺍ ﺍﻟﻐﺮﺽ ﺍﺭﺗﺄﻳﻨﺎ ﺗﻘﺴﻴﻢ ﻫﺬﺍ ﺍﳌﻮﺿﻮﻉ ﺇﱃ ﺃﺑﻮﺍﺏ ﻣﺘﺒﺎﻳﻨﺔ‪ ،‬ﻳﻜﻮﻥ ﻓﻴﻬﺎ‬
‫ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ ﻣﻮﺟﻪ ﻟﻠﻤﺴـﻴـّﺮﻳﻦ ﻏﲑ ﺍﳌﺘﺨﺼﺼﲔ ﰲ ﻣﻴﺪﺍﻥ ﺍﻟﺘﻨﺒﺆ ﻭﺑﺎﳋﺼﻮﺹ‬
‫ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ‪ ،‬ﻛﻮﻥ ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﻣﻦ ﺍﻟﺘﻘﻨﻴﺎﺕ ﻭﺍﻟﻨﻤﺎﺫﺝ ﻻ ﲢﺘﺎﺝ ﺇﱃ ﺇﳌﺎﻡ‬
‫ﻭﺍﺳﻊ ﲜﻮﺍﻧﺐ ﺍﻟﻨﻈﺮﻳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﳋﺎﺻﺔ ﲟﻮﺿﻮﻉ ﺍﻟﺪﺭﺍﺳﺔ ﻭﻻ ﺇﱃ ﺩﺭﺍﻳﺔ ﻛﺒﲑﺓ‬
‫ﲟﻮﺿﻮﻉ ﻭﻋﻤﻠﻴﺎﺕ ﺍﻟﺘﻨﺒﺆ‪ ،‬ﻟﺒﺴﺎﻃﺔ ﺗﺮﻛﻴﺒﻬﺎ‪ ،‬ﻭﻣﻌﻘﻮﻟﻴﺔ ﺣﻠﻬﺎ ﻭﺳﻬﻮﻟﺔ ﺗﻔﺴﲑ‬
‫ﻧﺘﺎﺋﺠﻬﺎ‪ ،‬ﻛﻤﺎ ﺃ‪‬ﺎ ﻻ ﲢﺘﺎﺝ ﺇﱃ ﺇﻣﻜﺎﻧﻴﺎﺕ ﻛﺒﲑﺓ ﻣﺜﻞ ﺃﺟﻬﺰﺓ ﺍﻟﻜﻤﺒﻴﻮﺗﺮ ﺍﳌﺘﻄﻮﺭﺓ‬
‫ﳌﺨﺘﻠﻒ ﺍﻟﻌﻤﻠﻴﺎﺕ ﺍﳊﺴﺎﺑﻴﺔ‪.‬‬
‫ﺑﻴﻨﻤﺎ ﺗﻘﻨﻴﺎﺕ ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ ﺗﻜﻮﻥ ﰲ ﺍﻟﻐﺎﻟﺐ ﻣﻮﺟﻬﺔ ﻟﻠﻤﺨﺘﺼﲔ ﰲ ﻫﺬﺍ‬
‫ﺍ‪‬ﺎﻝ‪ ،‬ﻭﻫﻲ ﺗﻌﺘﻤﺪ ﰲ ﺍﻷﺳﺎﺱ ﻋﻠﻰ ﺩﺭﺍﺳﺔ ﺍﻟﺴﻠﻮﻙ ﺍﳌﺎﺿﻲ ﻟﻠﻤﺘﻐﲑ ﺍﳌﺮﺍﺩ ﺍﻟﺘﻨﺒﺆ ﺑﻪ‬
‫ﻭﻓﻖ ﺃﺳﺲ ﺇﺣﺼﺎﺋﻴﺔ ﻗﻮﻳﺔ‪ ،‬ﰒ ﺍﻟﻘﻴﺎﻡ ﺑﻌﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﰲ ﺍﳌﺴﺘﻘﺒﻞ ﻋﻠﻰ ﺃﺳﺎﺱ ﺫﻟﻚ‪.‬‬
‫ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﻣﻦ ﺍﻟﺘﻘﻨﻴﺎﺕ ﲢﺘﺎﺝ ﺇﱃ ﺇﻣﻜﺎﻧﻴﺎﺕ ﻣﻌﺘﱪﺓ ﻟﻠﻘﻴﺎﻡ ‪‬ﺬﺍ ﺍﻟﺪﻭﺭ ﺍﻟﻔﻌﺎﻝ‪.‬‬
‫ﻟﻘﺪ ﰎ ﺇﺩﺭﺍﺝ ﺿﻤﻦ ﻫﺬﺍ ﺍﻟﻔﻀﻞ ﻧﺒﺬﻩ ﺑﺴﻴﻄﺔ ﻋﻦ ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ‬
‫ﺍﻟﺸﻌﺎﻋﻴﺔ )‪ (VAR‬ﻭﺗﻄﺒﻴﻘﺎ‪‬ﺎ ﺍﻟﻌﻤﻠﻴﺔ ﺇﺿﺎﻓﺔ ﺇﱃ ﺟﺪﻳﺪ ﺍﺧﺘﺒﺎﺭﺍﺕ ﺍﻻﺳﺘﻘﺮﺍﺭﻳﺔ ﺃﻭ‬
‫ﺳﻜﻮﻥ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪ .‬ﻟﻘﺪ ﰎ ﺍﺳﺘﻌﺮﺍﺽ ﺍﺧﺘﺒﺎﺭ ﺩﻳﻜﻲ ﻭﻓﻮﻟﺮ‬
‫)‪ (Dicky-Fuller‬ﻣﻊ ﺗﺒﻴﺎﻥ ﻛﻴﻔﻴﺔ ﺗﻄﺒﻴﻖ ﳐﺘﻠﻒ ﺃﺷﻜﺎﻟﻪ ﺑﺎﻻﺳﺘﻌﺎﻧﺔ ﺑﱪﻧﺎﻣﺞ‬
‫‪ Eviews‬ﻟﻠﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ‪.‬‬
‫ﻭﰲ ﺍﳋﺘﺎﻡ ﻭﺿﻤﻦ ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ ﺩﺍﺋﻤﺎﹰ ﻭﰲ ﺍﻟﻔﺼﻞ ﺍﻟﺴﺎﺩﺱ ﻭﺍﻷﺧﲑ‪ ،‬ﻧﻠﻘﻲ‬
‫ﻧﻈﺮﺓ ﻋﻠﻰ ﻛﻴﻔﻴﺔ ﺍﻟﺘﻨﺒﺆ ﻭﻓﻖ ﳕﺎﺫﺝ ﺑﻮﻛﺲ ‪ -‬ﺟﻨﻜﻴﻨﺰ )‪ (Box-Jenkins‬ﻣﻊ‬
‫ﺍﺳﺘﻌﺮﺍﺽ ﺍﻟﻮﺳﺎﺋﻞ ﺍﻹﺿﺎﻓﻴﺔ ﻟﻘﻴﺎﺱ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ‪ ،‬ﻭﻫﻲ ﺻﺎﳊﺔ ﻟﺘﻘﻨﻴﺎﺕ ﻭﳕﺎﺫﺝ‬
‫ﺍﻟﺒﺎﺑﲔ ﺍﻷﻭﻝ ﻭﺍﻟﺜﺎﱐ ﻋﻠﻰ ﺍﻟﺴﻮﺍﺀ‪ .‬ﻛﻤﺎ ﻧﺬﻛﱞﺮ ﺑﻮﺟﻮﺩ ﻧﻮﻋﲔ ﻣﻦ ﺍﻟﺘﻨﺒﺆ‪ ،‬ﺍﻷﻭﻝ‬
‫ﺍﺧﺘﺒﺎﺭﻱ ﻭﺍﻟﺜﺎﱐ ﻋﻤﻠﻲ‪ ،‬ﻳ‪‬ﻌﺘﻤﺪ ﻋﻠﻴﻪ ﰲ ﺍﲣﺎﺫ ﺍﻟﻘﺮﺍﺭﺍﺕ ﺍﳌﺴﺘﻘﺒﻠﻴﺔ‪.‬‬

‫‪12‬‬
 ‫ﺍﻟﻔﺼﻞ ﺍﻷﻭﻝ‬

‫ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺘﻨﺒﺆﻳﺔ‬

‫‪ -1‬ﺗﻌﺎﺭﻳﻒ‪:‬‬
‫ﺗ‪‬ﻌﺘﱪ ﳕﺎﺫﺝ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ )‪ (Econometric models‬ﻭﺳﻴﻠﺔ ﺫﺍﺕ‬
‫ﺃﳘﻴﺔ ﺑﺎﻟﻐﺔ ﰲ ﺗﻔﺴﲑ ﺑﻌﺾ ﺍﻟﻈﻮﺍﻫﺮ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻭﺍﻟﺘﻨﺒﺆ ﺑﺴﻠﻮﻛﻬﺎ ﺍﳌﺴﺘﻘﺒﻠﻲ‬
‫ﻷﻏﺮﺍﺽﹺ ﺃﳘﻬﺎ ﺍﻟﱪﳎﺔ ﻭﺍﻟﺘﺨﻄﻴﻂ ﺍﻻﻗﺘﺼﺎﺩﻱ‪ .‬ﻓﻬﻲ ﻋﺒﺎﺭﺓ ﻋﻦ ﻣﻌﺎﺩﻟﺔ ﺃﻭ ﳎﻤﻮﻋﺔ‬
‫ﻣﻌﺎﺩﻻﺕ ﺗﺘﺸﻜﻞ ﻣﻦ ﻣﺘﻐﲑﺍﺕ ﺩﺍﺧﻠﻴﺔ )ﺗﺎﺑﻌﺔ( ﻭﺃﺧﺮﻯ ﺧﺎﺭﺟﻴﺔ )ﻣﺴﺘﻘﻠﺔ(‬
‫ﺑﺎﻹﺿﺎﻓﺔ ﺇﱃ ﳎﻤﻮﻋﺔ ﻣﻌﻠﻤﺎﺕ ﻭﻣﻘﺎﺩﻳﺮ ﻋﺸﻮﺍﺋﻴﺔ‪ ،‬ﻭﲤﺜﻞ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻻﺕ ﻧﻈﺎﻣﺎﹰ‬
‫ﻛﺎﻣﻼﹰ ﻟﺘﺸﺒﻴﻪ ﳐﺘﻠﻒ ﻧﺸﺎﻃﺎﺕ ﺍﻻﻗﺘﺼﺎﺩ ﺍﻟﻮﻃﲏ‪.‬‬
‫ﺃﺧﺬﺕ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﻗﺴﻄﺎ ﻭﺍﻓﺮ‪‬ﺍ ﻣﻦ ﺍﻟﺪﺭﺍﺳﺔ ﻭﺍﻻﻫﺘﻤﺎﻡ ﻧﻈﺮﺍﹰ ﻻﺳﺘﻌﻤﺎﻻ‪‬ﺎ‬
‫ﺍﻟﻮﺍﺳﻌﺔ ﻭﺍﳌﺘﻌﺪﺩﺓ ﰲ ﺍﻟﺒﻠـﺪﺍﻥ ﺍﳌﺘﻄﻮﺭﺓ ﺍﻗﺘﺼﺎﺩﻳﺎ ﺳﻮﺍﺀً ﻣﻨﻬﺎ ﺍﻟﻐﺮﺑﻴﺔ )ﺍﻟﻠﻴﱪﺍﻟﻴﺔ( ﺃﻭ‬
‫ﺍﻟﺸﺮﻗﻴﺔ )ﺍﻻﺷﺘﺮﺍﻛﻴﺔ ﺳﺎﺑﻘﺎﹰ(‪ ،‬ﺣﻴﺚ ﺍﺳﺘ‪‬ﻌﻤﻠﺖ ﻛﺜﲑ‪‬ﺍ ﰲ ﺍﻹﲢﺎﺩ ﺍﻟﺴﻮﻓﻴﱵ ﻟﺘﻤﻜﻴﻨﻪ‬
‫ﻣﻦ ﺭﺻﺪ ﺍﻟﺘﻐﲑﺍﺕ ﺍﻟﻜﻠﻴﺔ ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﻟﻮﺿﻊ ﺧﻄﻄﻪ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﻟﻘﺼﲑﺓ‬
‫ﻭﺍﳌﺘﻮﺳﻄﺔ ﺍﻷﺟﻞ‪ ،‬ﰒ ﺗﻠﺘﻪ ﺍﻟﻮﻻﻳﺎﺕ ﺍﳌﺘﺤﺪﺓ ﺍﻷﻣﲑﻛﻴﺔ ﻗﺒﻞ ﻛﻞ ﻣﻦ ﺃﳌﺎﻧﻴﺎ‪،‬‬
‫ﺑﺮﻳﻄﺎﻧﻴﺎ ﻭﻓﺮﻧﺴﺎ ﰲ ﻣﺮﺣﻠﺔ ﻣﺘﺄﺧﺮﺓ ﺃﺛﻨﺎﺀ ﻭﺿﻌﻬﺎ ﳌﺨﻄﻂ ـ ﻣﻮﱐ ـ ‪Monnet‬‬
‫‪1‬‬

‫ﻭﻛﺬﺍ ﺍﺳﺘﺨﺪﺍﻡ ﺍﻟﻨﻤﺎﺫﺝ ﺍﶈﺎﺳﺒﻴﺔ ﰲ ﺑﺪﺍﻳﺔ ﺍﳋﻤﺴﻴﻨﺎﺕ ﰒ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ‬

‫ﺍﻟﻜﻠﻴﺔ ﻓﻴﻤﺎ ﺑﻌﺪ ﻧﻘﻼﹰ ﻋﻦ ﺍﻟﻮﻻﻳﺎﺕ ﺍﳌﺘﺤﺪﺓ ﺍﻷﻣﺮﻳﻜﻴﺔ ﰲ ﺑﺪﺍﻳﺔ ﺍﻟﺴﺒﻌﻴﻨﺎﺕ‪.2‬‬

‫‪1- Jean Omer Gabriel Monnet‬‬

‫‪2 - Les modèles appliqués de la macroéconomie, Anne Epaulard, Dunod, p19.‬‬

‫‪13‬‬
 ‫ﻋﻤﻠﻴﺔ ﺑﻨﺎﺀ ﳕﻮﺫﺝ ﻗﻴﺎﺱ ﺍﻗﺘﺼﺎﺩﻱ‪:‬‬
‫ﻳﺘﻢ ﺑﻨﺎﺀ ﳕﻮﺫﺝ ﻗﻴﺎﺱ ﺍﻗﺘﺼﺎﺩﻱ ﺳﻮﺍﺀ ﳍﺪﻑ ﺗﻌﻠﻴﻤﻲ ﺃﻭ ﻷﻏﺮﺍﺽ ﻋﻠﻤﻴﺔ‬
‫ﲝﺜﻴﺔ ﻟﻠﺘﻨﺒﺆ ﻣﻦ ﺟﻬﺔ ﺃﻭ ﻟﺘﺤﻠﻴﻞ ﺍﻟﺴﻴﺎﺳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻭﺍﲣﺎﺫ ﺍﻟﻘﺮﺍﺭ ﻣﻦ ﺟﻬﺔ ﺛﺎﻧﻴﺔ‪،‬‬
‫ﻭﺫﻟﻚ ﻣﻦ ﺧﻼﻝ ﺍﻻﺳﺘﻌﺎﻧﺔ ﺑﺎﻟﻌﻠﻮﻡ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ .1‬ﻋﻠﻢ ﺍﻻﻗﺘﺼﺎﺩ‪ :‬ﻳﻔﻴﺪ ﰲ ﻭﺿﻊ ﻫﻴﻜﻞ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﻨﻈﺮﻱ ﻣﻦ ﺧﻼﻝ ﺍﻟﻨﻈﺮﻳﺔ‬
‫ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﻟﱵ ﳝﻜﻦ ﺗﻌﺮﻳﻔﻬﺎ ﻋﻠﻰ ﺃ‪‬ﺎ ﻣﺒﺪﺃ ﺃﻭ ﳎﻤﻮﻋﺔ ﻣﺒﺎﺩﺉ ﻣﻌﻘﻮﻟﺔ ﻭﻣﺘﻔﻖ‬
‫ﻋﻠﻴﻬﺎ ﻟﺸﺮﺡ ﺃﻭ ﺗﻔﺴﲑ ﻇﺎﻫﺮﺓ ﺍﻗﺘﺼﺎﺩﻳﺔ ﻛﺎﻧﺖ ﺃﻭ ﲡﺎﺭﻳﺔ‪ ،‬ﻣﺎﻟﻴﺔ ‪...‬ﺍﱁ‪.‬‬
‫ﰲ ﻫﺬﺍ ﺍﻟﺴﻴﺎﻕ‪ ،‬ﻭﻟﻔﻬﻢ ﺍﻟﻈﻮﺍﻫﺮ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻳﻘﻮﻡ ﺍﳌﻨﻈﺮ ﺑﺎﻟﺒﺤﺚ ﻋﻦ‬
‫ﺍﻟﻘﻮﺍﻧﲔ ﺍﻟﱵ ﺗﺴﻴ‪‬ﺮﻫﺎ‪ ،‬ﻭﻣﻨﻪ ﻣﻌﺮﻓﺔ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﱵ ﲢﺪﺩ ﺍﲡﺎﻩ ﻣﺴﺎﺭﻫﺎ ﺍﳌﺴﺘﻘﺒﻠﻲ‪.‬‬
‫‪ .2‬ﺍﻟﺮﻳﺎﺿﻴﺎﺕ‪ :‬ﻳﺘﻤﺜﻞ ﺩﻭﺭﻫﺎ ﰲ ﺻﻴﺎﻏﺔ ﺍﻟﻨﻈﺮﻳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﰲ ﻗﺎﻟﺐ ﺭﻳﺎﺿﻲ‪ ،‬ﰲ‬
‫ﺷﻜﻞ ﻣﻌﺎﺩﻟﺔ ﺃﻭ ﳎﻤﻮﻋﺔ ﻣﻌﺎﺩﻻﺕ‪ ،‬ﺳﻮﺍﺀً ﻛﺎﻧﺖ ﺳﻠﻮﻛﻴﺔ‪ ،‬ﺗﻌﺮﻳﻔﻴﺔ ﺃﻭ ﺗﻮﺍﺯﻧﻴﺔ‪.‬‬
‫ﻛﻤﺎ ﻳﻔﻴﺪ ﻫﺬﺍ ﺍﻟﻌﻠﻢ ﰲ ﻣﺮﺣﻠـﺔ ﻻﺣﻘﺔ ﻭﺇﺿﺎﻓﺔ ﺇﱃ ﺍﻟﻌﻤﻠﻴﺎﺕ ﺍﻟﺮﻳﺎﺿﻴﺔ ﺍﳌﺨﺘﻠﻔﺔ‬
‫ﰲ ﺍﻟﺒﺤﺚ ﰲ ﺧﺼﺎﺋﺺ ﺍﻟﻨﻤﻮﺫﺝ‪.‬‬
‫‪ .1-2‬ﺍﳌﻌﺎﺩﻟﺔ ﺍﻟﺴﻠﻮﻛﻴﺔ‪ :‬ﻫﻲ ﺗﺮﲨﺔ ﺭﻳﺎﺿﻴﺔ ﻟﻌﺎﺩﺍﺕ ﺍﻟﻌﻤﻼﺀ ﺍﻻﻗﺘﺼﺎﺩﻳﲔ ﻣﻦ‬
‫ﻣﻨﺘﺠﲔ )ﻋﻼﻗﺎﺕ ﺗﻘﻨﻴﺔ( ﻭﻣﺴﺘﺜﻤﺮﻳﻦ ﻭﻣﺴﺘﻬﻠﻜﲔ‪...‬ﺍﱁ‪ ،‬ﺣﻴﺚ ﺗﻔﺴﺮ ﻇﺎﻫﺮﺓ‬
‫ﺍﻗﺘﺼﺎﺩﻳﺔ ﺑﺪﻻﻟﺔ ﻣﺘﻐﲑ ﺃﻭ ﳎﻤﻮﻋﺔ ﻣﻦ ﺍﳌﺘﻐﲑﺍﺕ‪ ،‬ﻭﻟﻜﻮ‪‬ﺎ ﻛﺬﻟﻚ‪ ،‬ﻓﻬﻲ ﻏﲑ‬
‫ﺻﺤﻴﺤﺔ ﺑﺎﻟﺘﻌﺮﻳﻒ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﻓﻬﻲ ﻗﺎﺑﻠﺔ ﻟﻼﺧﺘﺒﺎﺭ‪.‬‬
‫‪Ct = a + bYt + Ut‬‬ ‫)‪(1.1‬‬ ‫ﻣﺜﺎﻝ ‪:1‬‬
‫ﺣﻴﺚ ‪ C t‬ﻭ ‪ Yt‬ﲤﺜﻼﻥ ﺍﻻﺳﺘﻬﻼﻙ ﻛﻤﺘﻐﲑﹴ ﺗﺎﺑﻊ ﻭﺍﻟﺪﺧﻞ ﻛﻤﺘﻐﲑﹴ ﻣﺴﺘﻘﻞ‪ 3‬ﻋﻠﻰ‬

‫‪ -3‬ﻳﺸﻤﻞ ﺍﳌﺘﻐﲑ ﺍﳌﺴﺘﻘﻞ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﺘﺎﺭﳜﻴﺔ )ﻣﺘﻐﲑ ﺃﻭ ﻣﺘﻐﲑﺍﺕ ﻣﺆﺧﺮﺓ ﺑﻔﺘﺮﺓ ﺃﻭ ﻓﺘﺮﺍﺕ‬
‫ﺯﻣﻨﻴﺔ "‪ ("Lagged dependent variables‬ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﻘﺮﺍﺭﻳﺔ )‪(decision variables‬‬
‫ﻭﻣﺘﻐﲑﺍﺕ ﺍﻟﺴﻴﺎﺳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺇﺿﺎﻓﺔﹰ ﺇﱃ ﻣﺘﻐﲑﺍﺕ ﺍﶈﻴﻂ ﺍﻟﺪﻭﱄ‪.‬‬

‫‪14‬‬
‫ﺍﻟﺘﺮﺗﻴﺐ‪ ،‬ﻭﻓﻖ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ‪ ،4‬ﺑﻴﻨﻤﺎ ‪ U t‬ﲤﺜﻞ ﺧﻄﺄﹰ ﻋﺸﻮﺍﺋﻴﺎﹰ ﻳﻌﱪ ﻋﻦ ﺍﳌﺘﻐﲑﺍﺕ‬
‫ﺍﻟﱵ ﻻ ﳝﻜﻦ ﻗﻴﺎﺳﻬﺎ ﻛﺎﻟﺬﹼﻭﻕ ﻣﺜـﻼﹰ ﺃﻭ ﺍﻷﺧﻄﺎﺀ ﺍﻟﱵ ﺗﻘﻊ ﺃﺛﻨﺎﺀ ﲨﻊ‪ ،‬ﻭﺗـﺪﻭﻳﻦ‬
‫ﻭﻃﺒﻊ ﺍﳌﻌﻠﻮﻣﺎﺕ‪ ،‬ﺑﻴﻨﻤﺎ )‪ (t‬ﲤﺜﻞ ﺩﻟﻴﻞ ﺍﻟﺰﻣﻦ ﺣﻴﺚ ﻳﺄﺧﺬ ﺍﻟﻘﻴ‪‬ﻢ )‪ (T،..،2،1‬ﻭﻫﺬﺍ‬
‫ﺍﻷﺧﲑ ﻳﻌﱪ ﻋﻦ ﺁﺧﺮ ﻣﺸﺎﻫﺪﺓ ﻣﺘﻮﻓﺮﺓ ﺃﻭ ﻣﺴﺘﻌﻤﻠﺔ ﰲ ﺍﻟﻌﻴﻨﺔ ﺍﳌﺪﺭﻭﺳﺔ‪ ،‬ﺃﻣﺎ ‪a‬‬
‫ﻭ ‪ b‬ﲤﺜﻞ ﺛﺎﺑﱵ ﺍﻟﺪﺍﻟﺔ ﺍﻟﱵ ﻳﺘﻢ ﺗﻘﺪﻳﺮﻫﺎ‪.‬‬
‫‪ -2-2‬ﺍﳌﻌﺎﺩﻟﺔ ﺍﻟﺘﻌﺮﻳﻔﻴﺔ‪ :‬ﻭﻣﻦ ﺧﻼﻝ ﺗﺴﻤﻴﺘﻬﺎ ﺗﻈﻬﺮ ﺑﺄ‪‬ﺎ ﺗﻔﻴﺪ ﰲ ﺍﻟﺘﻌﺮﻳﻒ‬
‫ﺍﻟﺮﻳﺎﺿﻲ ﳌﺘﻐﲑ ﺍﻗﺘﺼﺎﺩﻱ ﻣﺜﻼﱠً‪ ،‬ﻓﻬﻲ ﺇﺫﺍ ﺻﺤﻴﺤﺔ ﺑﺎﻟﺘﻌﺮﻳﻒ ﻭﻓﻖ ﻧﻈﺎﻡ ﳏﺎﺳﱯ‬
‫ﻣﻌﲔ ﻭﻏﲑ ﻗﺎﺑﻠﺔ ﻟﻼﺧﺘﺒﺎﺭ‪.‬‬
‫‪Y =C + I‬‬
‫‪t‬‬ ‫‪t‬‬ ‫‪t‬‬
‫)‪(1.2‬‬ ‫ﻣﺜﺎﻝ ‪:2‬‬

‫‪Ct‬‬ ‫ﻓﻬﻲ ﺑﺬﻟﻚ ﲤﺜﻞ ﻣﻌﺎﺩﻟﺔ ﳏﺎﺳﺒﻴﺔ‪ ،‬ﺗﻘﺴﻢ ﺍﻟﺪﺧﻞ ﺇﱃ ﻣﺮﻛﺒﱵ ﺍﻻﺳﺘﻬﻼﻙ‬
‫ﻭﺍﻻﺳﺘﺜﻤﺎﺭ ‪ I t‬ﰲ ﺍﻟﻔﺘﺮﺓ )‪.(t‬‬
‫‪ .3-2‬ﺍﳌﻌﺎﺩﻟﺔ ﺍﻟﺘﻮﺍﺯﻧﻴﺔ‪ :‬ﺗﻌﺘﱪ ﻛﻤﻌﺎﺩﻟﺔ ﺷﺮﻁ ﺍﻗﺘﺼﺎﺩﻱ ﻟﺘﻤﺜﻞ ﻭﺿﻌﺎﹰ ﺗﻮﺍﺯﻳﹰﺎ‬
‫ﻳﺘﺴﺎﻭﻯ ﻓﻴﻪ ﻣﺜﻼﹰ ﺍﻟﻌﺮﺽ ‪ S t‬ﻭﺍﻟﻄﻠﺐ ‪ Dt‬ﻋﻠﻰ ﺳﻠﻌﺔ ﻣﺎ ﰲ ﺳﻮﻕ ﻭﻓﺘﺮﺓ ﺯﻣﻨﻴﺔ‬
‫ﻣﻌﻴﻨﺘﲔ‪ .‬ﺗﺴﺎﻫﻢ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻻﺕ ﺑﺸﻜﻞ ﻛﺒﲑ ﰲ ﻋﻤﻠﻴﺔ ﺣﻞ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻵﻧﻴﺔ‬
‫ﻭﺗﺴﻬﻴﻞ ﻋﻤﻠﻴﺔ ﺗﻌﺮﻳﻒ ﺍﻟﻨﻤﻮﺫﺝ )‪(Simultaneous Model Identification‬‬
‫‪St = Dt‬‬ ‫)‪(1.3‬‬ ‫‪:3‬‬ ‫ﻣﺜﺎﻝ‬
‫‪ .3‬ﺍﻹﺣﺼﺎﺀ‪ :‬ﻳﺘﻢ ﻣﻦ ﺧﻼﻟﻪ ﺗﻮﻓﲑ ﻗﺼﺪ ﺍﻻﺳﺘﻐﻼﻝ ﻟﻠﻤﻌﻄﻴﺎﺕ ﺍﳌﻴﺪﺍﻧﻴﺔ ﺑﻌﺪ‬
‫ﺻﻘﻠﻬﺎ ﻭﺗﺒﺴﻴﻄﻬﺎ ﻟﻼﺳﺘﻌﻤﺎﻝ ﻭﺍﻟﺘﺤﻠﻴﻞ‪ .‬ﻛﻤﺎ ﺗﻮﻓﺮ ﺃﺟﻬﺰﺗﻪ ﺍﳌﺨﺘﺼﺔ ﺍﳌﻌﻄﻴﺎﺕ‬
‫ﺍﻟﻀﺮﻭﺭﻳﺔ ﻟﻌﻤﻠﻴﺔ ﺍﻟﻨﻤﺬﺟﺔ ﻭﻛﺬﺍ ﺃﺩﻭﺍﺕ ﺍﻻﺧﺘﺒﺎﺭ ﺍﻟﻼﺯﻣﺔ‪ ،‬ﻣﺜﻞ ﺇﺣﺼﺎﺋﻴﺎﺕ "‪t-‬‬
‫‪ ،"Student‬ﻓﻴﺸﺮ "‪" Fisher‬ﻭﻛﻲ ﻣﺮﺑﻊ) ‪...( c 2‬ﺍﱁ‪.‬‬

‫‪ - 4‬ﳕﻮﺫﺝ ﻛﻴﻨﺰ ﻟﻼﺳﺘﻬﻼﻙ‬

‫‪15‬‬
 ‫‪ -2‬ﺃﻧﻮﺍﻉ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ‪:‬‬
‫ﳝﻜﻦ ﻭﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ﺗﻘﺴﻴﻢ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﺇﱃ ﻧﻮﻋﲔ ﻣﺘﻤﻴﺰﻳﻦ ﺷﺎﺋﻌﲔ ﰲ‬
‫ﳎﺎﻝ ﺍﻟﺘﻄﺒﻴﻘﺎﺕ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻳﺘﻤﺜﻼﻥ ﰲ‪:‬‬
‫‪ -1-2‬ﳕﺎﺫﺝ ﺍﻟﱪﳎﺔ ﺍﻟﺮﻳﺎﺿﻴﺔ ‪Mathematical programming programs‬‬
‫ﺍﻟﱵ ﺗﺘﻤﺜﻞ ﰲ ﻋﻤﻠﻴﺔ ﺗﺼﻤﻴﻢ ﺃﻭ ﺻﻴﺎﻏﺔ ﳕﻮﺫﺝ ﺫﻭ ﻋﻼﻗﺎﺕ ﺭﻳﺎﺿﻴﺔ ﻭﻣﻨﻄﻘﻴﺔ‬
‫ﺍﺳﺘﺠﺎﺑﺔ ﻟﻀﺮﻭﺭﺍﺕ ﺍﻟﺒﺤﻮﺙ ﺍﳌﻴﺪﺍﻧﻴﺔ‪ .‬ﺃﺑﺴﻂ ﻫﺬﻩ ﺍﻟﺘﻘﻨﻴﺎﺕ‪ ،‬ﳕﺎﺫﺝ ﺍﻟﱪﳎﺔ ﺍﳋﻄﻴﺔ‬
‫ﻭﺍﻟﺪﻳﻨﺎﻣﻴﻜﻴﺔ‪.‬‬
‫‪ -2-2‬ﳕﺎﺫﺝ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ ‪Econometric models‬‬
‫ﻓﺈﺫﺍ ﺍﻫﺘﻢ ﺍﻟﻨﻮﻉ ﺍﻷﻭﻝ ﺑﺎﻟﺒﺤﺚ ﻋﻦ ﺑﺈﳚﺎﺩ ﺍﳊﻞ ﺍﻷﻣﺜﻞ ﻟﺒﻌﺾ‬
‫ﺍﻹﺷﻜﺎﻻﺕ ﺍﳌﺘﻤﺜﻠﺔ ﰲ ﺗﻌﻈﻴﻢ )‪ (Maximize‬ﺃﻭ ﺗﺪﻧﻴﺔ )‪ (Minimize‬ﺩﺍﻟﺔ‬
‫ﺍﳍﺪﻑ ﻭﻓﻖ ﻗﻴﻮﺩ ﻣﻌﻠﻮﻣﺔ‪ ،‬ﻓﺈﻥ ﺍﻟﻨﻮﻉ ﺍﻟﺜﺎﱐ ﻳﻔـﻴﺪ ﰲ ﻣﻌﺮﻓﺔ ﻭﺭﺻﺪ ﺍﻟﺴﻠﻮﻙ‬
‫ﺍﳌﺎﺿﻲ ﻟﻠﻤﺘﻐﲑﺍﺕ‪ ،‬ﰒ ﺍﻟﺘﻨﺒﺆ ﺑﺴﻠﻮﻛﻬﺎ ﺍﳌﺴﺘﻘﺒﻠﻲ‪ .‬ﻛﻤﺎ ﻳﻔﻴﺪ ﺃﻳﻀﺎﹰ ﰲ ﲢﻠﻴﻞ‬
‫ﺍﻟﺴﻴﺎﺳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻟﻠﺪﻭﻟﺔ ﻭﺍﲣﺎﺫ ﺍﻟﻘﺮﺍﺭ ﻋﻠﻰ ﺍﳌﺴﺘﻮﻳﲔ ﺍﳉﺰﺋﻲ ﺃﻭ ﺍﻟﻜﻠﻲ‬
‫)‪.(Policy Analysis & Decision making‬‬

‫ﻳﻘﺘﺼﺮ ﺍﻻﻫﺘﻤﺎﻡ ﺍﻵﻥ ﻋﻠﻰ ﳏﺎﻭﻟﺔ ﺍﻟﺘﻄﺮﻕ ﺇﱃ ﳕﺎﺫﺝ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ‬

‫ﻣﻌﺘﻤﺪﻳﻦ ﻋﻠﻰ ﻣﻌﻄﻴﺎﺕ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﱵ ﺗﻌﺮﻑ ﻋﻠﻰ ﺃﺳﺎﺱ ﺃ‪‬ﺎ ﻣﻌﻠﻮﻣﺎﺕ ﺭﻗﻤﻴﺔ‬
‫ﺧﺎﺻﺔ ﺑﺎﳌﺘﻐﲑﺍﺕ ﺃﻭ ﺍﻟﻈﻮﺍﻫﺮ ﺍﳌﺪﺭﻭﺳﺔ ﺍﻟﱵ ﺗﻜﻮﻥ ﰲ ﺍﻟﻐﺎﻟﺐ ﺫﺍﺕ ﻃﺎﺑﻊ ﺍﻗﺘﺼﺎﺩﻱ‬
‫ﺍﺟﺘﻤﺎﻋﻲ∙ ﺗﻐﻄﻲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻓﺘﺮﺓ ﳏﺪﺩﺓ ﻣﻦ ﺍﻟﺰﻣﻦ ﰲ ﺷﻜﻞ ﻣﻌﻄﻴﺎﺕ ﻗﺪ ﺗﻜﻮﻥ‬
‫ﻗﺎﺑﻠﺔ ﻟﻠﻤﻘﺎﺭﻧﺔ ﻛﺎﻟﺴﻨﺔ‪ ،‬ﺍﻟﻔﺼﻞ‪ ،‬ﺍﻟﺸﻬﺮ ﺍﻷﺳﺒﻮﻉ ﻭﺍﻟﻴﻮﻡ ∙∙∙ﺍﱁ‪.‬‬
‫ﰲ ﻫﺬﺍ ﺍﻟﺴﻴﺎﻕ‪ ،‬ﳝﻜﻦ ﺍﻹﺷﺎﺭﺓ ﳑﺎ ﻳﺘﺒﲔ ﰲ ﺍﻟﺸﻜﻞ )‪ (1‬ﺃﺩﻧﺎﻩ‪ ،‬ﺃﻥ ﻋﻤﻠﻴﺔ‬
‫ﲢﻠﻴﻞ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺗﺴﺎﻫﻢ ‪ -‬ﺇﺿﺎﻓﺔ ﻟﻌﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ‪ -‬ﰲ ﲢﻀﲑ ﳐﻄﻄﺎﺕ‬
‫ﻭﺑﺮﺍﻣﺞ ﺗﻮﻗﻊ ﺣﺠﻢ ﺍﻟﻄﻠﺒﺎﺕ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﺆﺳﺴﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ‪ ،‬ﰲ ﻭﺿﻊ ﲢﺖ‬
‫ﺗﺼﺮﻑ ﺍﳌﺴﲑﻳﻦ ﺣﻘﺎﺋﻖ ﻛﻤﻴﺔ ﻋﻦ ﺍﻟﻮﺿﻊ ﺍﳊﻘﻴﻘﻲ ﻟﻠﺘﻨﻤﻴﺔ ﻭﺫﻟﻚ ﻟﺘﻔﺎﺩﻱ‬
‫ﺍﻟﻘﺮﺍﺭﺍﺕ ﺍﳋﺎﻃﺌﺔ∙‬

‫‪16‬‬
 ‫ﺍﻟﺸﻜﻞ )‪ :(1‬ﺃﻧﻮﺍﻉ ﳕﺎﺫﺝ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ ﻭﺃﻫﺪﺍﻓﻬﺎ‬

‫)‪(Regression models‬‬ ‫‪ -1-2‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ‪:‬‬

‫ﺗﺸﺮﺡ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﻣﺘﻐﲑ ﺗﺎﺑﻊ ﺑﻮﺍﺳﻄﺔ ﻣﺘﻐﲑ ﺃﻭ ﳎﻤﻮﻋﺔ ﻣﻦ ﺍﳌﺘﻐﲑﺍﺕ‬
‫ﺍﳌﺴﺘﻘﻠﺔ ﰲ ﺃﺑﺴﻂ ﺃﺷﻜﺎﳍﺎ‪ ،‬ﻭﻟﺘﻮﺿﻴﺢ ﺫﻟﻚ ﻧﺪﺭﺝ ﻛﻤﺜﺎﻝ ﳕﻮﺫﺝ ﺍﻻﺳﺘﻬﻼﻙ‬
‫ﺍﻟﻜﻨﺰﻱ )‪ (Keynesian consumption function‬ﺍﻟﺘﺎﱄ‪:‬‬
‫) ‪Ct = f (Yt‬‬ ‫)‪(1.4‬‬

‫ﻧﺸﲑ ﻫﻨﺎ‪ ،‬ﺇﱃ ﺃﻥ ﺍﻟﻨﻈﺮﻳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺗﻌﺠﺰ ﻋﻦ ﲢﺪﻳﺪ ﺷﻜﻞ ﺍﻟﺪﺍﻟﺔ‬

‫) ‪f (.‬‬
‫ﻭﻛﺬﺍ ﻋﺪﺩ ﻣﻌﺎﺩﻻﺕ ﺍﻟﻨﻤﻮﺫﺝ ﻳﺘﻢ ﺍﻟﺘﻐﻠﺐ ﻋﻠﻰ ﺍﻹﺷﻜﺎﻝ ﺍﻷﻭﻝ ﻛﻤﺎ ﻫﻮ ﺍﳊﺎﻝ‬
‫ﻫﻨﺎ ﺑﻮﺍﺳﻄﺔ ﺭﺳﻢ ﺑﻴﺎﱐ ﻣﺰﺩﻭﺝ ﺑﲔ ﺍﳌﺘﻐﲑ ﺍﻟﺘﺎﺑﻊ ﻭﺍﳌﺴﺘﻘﻞ ﺍﻵﻥ ﻭﺑﺎﻓﺘﺮﺍﺽ ﺃﻥ‬
‫ﺍﻟﻌﻼﻗﺔ ﺍﻟﱵ ﺗﺮﺑﻂ ﺑﲔ ﺍﻻﺳﺘﻬﻼﻙ ﻭﺍﻟﺪﺧﻞ ﺗﺘﻤﺜﻞ ﰲ ﺩﺍﻟﺔ ﺧﻄﻴﺔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪C t = a + b Y t + Ut‬‬

‫ﺣﻴﺚ ‪ C‬ﻭ ‪ Yt‬ﲤﺜﻼﻥ ﺍﳌﺘﻐﲑ ﺍﻟﺘﺎﺑﻊ ﻭﺍﳌﺴﺘﻘﻞ ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪ ،‬ﺃﻱ ﺍﳌﺘﺄﺛﺮ‬ ‫‪t‬‬

‫ﻭﺍﳌﺆﺛﺮ‪ ،‬ﺣﻴﺚ ﻳﺘﺤﺪﺩ ﺍﳌﺘﻐﲑ ﺍﻷﻭﻝ ﲟﻌﺮﻓﺔ ﺍﳌﺘﻐﲑ ﺍﻟﺜﺎﱐ‪ .‬ﺑﻴﻨﻤﺎ ‪ a‬ﻭ ‪ b‬ﲤﺜﻼﻥ‬
‫ﺍﻻﺳﺘﻬﻼﻙ ﺍﻟﺘﻠﻘﺎﺋﻲ ﻭﺍﳌﻴﻞ ﺍﳊﺪﻱ ﻟﻼﺳﺘﻬﻼﻙ ﻋﻠﻰ ﺍﻟﺘﻮﺍﱄ‪.‬‬

‫‪17‬‬
‫ﻛﻤﺎ ﳝﻜﻦ ﺃﻥ ﻳﺘﻜﻮﻥ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻻﳓﺪﺍﺭﻱ ﻣﻦ ﻣﻌﺎﺩﻟﺔ ﺃﻭ ﳎﻤﻮﻋﺔ‬
‫ﻣﻌﺎﺩﻻﺕ ﻗﺪ ﺗﺸﻜﻞ ﻓﻴﻤﺎ ﺑﻴﻨﻬﺎ ﳕﻮﺫﺝ ﺁﱐ‪(Simultaneous equations ،‬‬
‫)‪ model‬ﰲ ﺷﻜﻞ ﻣﻌﺎﺩﻻﺕ ﺳﻠﻮﻛﻴﺔ‪ ،‬ﻣﻌﺎﺩﻻﺕ ﺗﻌﺮﻳﻔﻴﺔ ﻭﺃﺧﺮﻱ ﺗﻮﺍﺯﻧﻴﺔ‪ .‬ﻓﺈﺫﺍ‬
‫ﻛﺎﻧﺖ ﺍﻟﺪﺭﺍﺳﺔ ﺫﺍﺕ ﺃﺑﻌﺎﺩ ﺍﻗﺘﺼﺎﺩﻳﺔ ﻛﻠﻴﺔ ﻳﻄﻠﻖ ﻋﻠﻰ ﺍﻟﺼﻴﺎﻏﺔ ﺍﻟﺮﻳﺎﺿﻴﺔ‬
‫ﺍﻟﺒﺴﻴﻄﺔ ‪‬ﻤﻮﻋﺔ ﺗﻠﻚ ﺍﻟﻈﻮﺍﻫﺮ ﺍﳌﻌﻘﺪﺓ ﺑﻨﻤﻮﺫﺝ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ ﺍﻟﻜﻠﻲ‬
‫)‪.(Macroeconometric model‬‬
‫ﻟﺘﻮﺿﻴﺢ ﺫﻟﻚ‪ ،‬ﻧﺪﺭﺝ ﺍﳉﺪﻭﻝ )‪ (1‬ﺍﻟﺘﺎﱄ ﺍﳌﺘﻌﻠﻖ ﺑﻌﻴﻨﺔ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ‬
‫ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﻟﻜﻠﻴﺔ ﺳﻮﺍﺀً ﻛﺎﻧﺖ ﺫﺍﺕ ﺃﺑﻌﺎﺩ ﺑﻴﺪﺍﻏﻮﺟﻴﺔ ﺃﻭ ﺗﻄﺒﻴﻘﻴﺔ ﺍﳋﺎﺻﺔ ﺑﺜﻼﺙ‬
‫ﺩﻭﻝ ﺻﻨﺎﻋﻴﺔ‪ ،‬ﺑﺮﻳﻄﺎﻧﻴﺎ )‪ ،(UK‬ﺍﻟﻮﻻﻳﺎﺕ ﺍﳌﺘﺤﺪﺓ ﺍﻷﻣﺮﻳﻜﻴﺔ )‪ (USA‬ﻭﻓﺮﻧﺴﺎ‬
‫)‪. (FR‬‬
‫‪5‬‬

‫ﺍﳉﺪﻭﻝ) ‪ :( 1‬ﳎﻤﻮﻋﺔ ﳕﺎﺫﺝ ﻗﻴﺎﺳﻴﺔ ﻛﻠﻴﺔ‪.‬‬

‫ﺍﻟﻨﻤﻮﺫﺝ‬ ‫ﺍﻟﺒﻠﺪ‬ ‫ﺍﳌﻌﺎﺩﻻﺕ‬ ‫ﺍﻟﻨﺎﺷـــﺮ‬
‫‪Treasury‬‬ ‫‪UK‬‬ ‫‪583‬‬ ‫& ‪Economic Trend‬‬
‫‪Economic Forecast‬‬
‫‪London Business School‬‬ ‫‪UK‬‬ ‫‪284‬‬ ‫‪Sunday Times‬‬

‫‪National Institute‬‬ ‫‪UK‬‬ ‫‪150‬‬ ‫‪National. Institute‬‬

‫‪Economic revue‬‬
‫‪Federal Reserve Board‬‬ ‫‪USA‬‬ ‫‪194‬‬ ‫‪----‬‬
‫‪Warton School of Pennsylvania‬‬ ‫‪USA‬‬ ‫‪207‬‬ ‫‪----‬‬
‫‪Copain‬‬ ‫‪FR‬‬ ‫‪350‬‬ ‫‪----‬‬
‫‪Metric‬‬ ‫‪FR‬‬ ‫‪1000‬‬ ‫‪----‬‬

‫ﻣﻦ ﺃﻫﺪﺍﻑ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ‪ ،‬ﺍﻟﺘﻨﺒﺆ‪ ،‬ﺍﻟﺬﻱ ﻳﻔﻴﺪ ﰲ ﲢﺪﻳﺪ ﻗﻴ‪‬ﻢ ﺍﳌﺘﻐﲑﺍﺕ ﺫﺍﺕ‬
‫ﺍﻷﳘﻴﺔ ﺑﺎﻟﻨﺴﺒﺔ ﳌﺘﺨﺬ ﺍﻟﻘﺮﺍﺭ ﻣﺴﺘﻘﺒﻠﻴﺎﹰ‪ .‬ﻧﺸﲑ ﻫﻨﺎ ﺇﱄ ﻭﺟﻮﺩ ﻧﻮﻋﲔ ﻣﻦ ﺍﻟﺘﻨﺒﺆ‪،‬‬

‫‪1) Topics in applied econometrics K.F.wallis, p.128‬‬ ‫‪ .5‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ ‪:‬‬

‫‪2) Comptabilité Nationale et Modèles Economiques, A. PICHOT, puf, 1988.‬‬

‫‪18‬‬
‫ﺍﻷﻭﻝ ﺗﺎﺭﳜﻲ ﺍﺧﺘﺒﺎﺭﻱ )‪ ،(Ex-post‬ﻳﺴﺘﺨﺪﻡ ﻛﺎﺧﺘﺒﺎﺭ ﳌﻌﺮﻓﺔ ﺩﻗﺔ ﺗﻨﺒﺆ ﺍﻟﻨﻤﻮﺫﺝ‬
‫ﺍﻟﺮﻳﺎﺿﻲ ﺍﳌﺼﻤﻢ ﻋﻦ ﻃﺮﻳﻖ ﻣﻘﺎﺭﻧﺔ ﺍﻟﻘﻴ‪‬ﻢ ﺍﳌﺸﺎﻫﺪﺓ ﻣﻊ ﺍﻟﺘﻨﺒﺆ ‪(Actual versus‬‬
‫)‪ ،Forecast‬ﻭﺍﻟﺜﺎﱐ ﺗﻨﺒﺆ ﻣﺴﺘﻘﺒﻠﻲ )‪ (Ex-ante‬ﻋﻤﻠﻲ ﻣﻴﺪﺍﱐ‪ ،‬ﻭﻳﺘﻢ ﺑﻌﺪ ﳒﺎﺡ‬
‫ﺍﻷﻭﻝ ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﺳﺎﺑﻘﺎ‪.‬‬
‫ﻛﻤﺎ ‪‬ﺪﻑ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﺃﻳﻀﺎﹰ‪ ،‬ﺇﱃ ﲢﻠﻴﻞ ﺍﻟﺴﻴﺎﺳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻭﺍﲣﺎﺫ‬
‫ﺍﻟﻘﺮﺍﺭ‪ ،‬ﺣﻴﺚ ﺗﻘـﻮﻡ ﺍﻟﺴﻠﻄﺔ ﺍﳌﻌﻨﻴﺔ ﺑﺎﲣﺎﺫ ﺍﻟﻘﺮﺍﺭ ﺳﻮﺍﺀً ﻛﺎﻧﺖ ﺍﻗﺘﺼﺎﺩﻳﺔ‪ ،‬ﲡﺎﺭﻳﺔ‪،‬‬
‫ﻣﺎﻟﻴﺔ ﺃﻭ ﻧﻘﺪﻳﺔ ﺃﻭ ﻏﲑﻫﺎ‪ ،‬ﺇﱃ ﺑﺘﺠﺮﻳﺐ ﺳﻴﺎﺳـﺘﻬﺎ ﺍﳌﺴﺘﻬﺪﻓﺔ ﻋﻦ ﻃﺮﻳﻖ ﺍﶈﺎﻛﺎﺓ‬
‫)‪ (simulation‬ﻣﻦ ﺧﻼﻝ ﺗﻮﺟﻴﻪ ﻭﲢﺪﻳﺪ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﻘﺮﺍﺭﻳﺔ ‪(Decision‬‬
‫)‪ ،variables‬ﻗﺒﻞ ﺗﻨﻔﻴﺬﻫﺎ ﻣﻴﺪﺍﻧﻴﺎﹰ ﻭﻣﻨﻪ ﲡﻨﺐ ﺍﻷﺿﺮﺍﺭ ﺍﳉـﺎﻧﺒﻴﺔ ﻏﻴــﺮ‬
‫ﺍﳌﺘﻮﻗﻌﺔ‪.‬‬
‫ﳍﺬﺍ ﻓﻘﺪ ﺍﺳﺘﺨﺪﻣﺖ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻜﻠﻴﺔ ﺑﺸﻜﻞ ﻭﺍﺳﻊ ﻣﻦ ﻃﺮﻑ ﻣﺘﺨﺬﻱ‬
‫ﺍﻟﻘﺮﺍﺭ ﻟﺘﻮﺟﻴﻪ ﺍﻟﺴﻴﺎﺳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻭﻛﺬﺍ ﲢﻘﻴﻖ ﺍﻟﺘﻨﺒﺆﺍﺕ ﻣﻦ ﻃﺮﻑ ﺍﳌﺆﺳﺴﺎﺕ‬
‫ﺍﳊﻜﻮﻣﻴﺔ ﻭﺍﻟﺪﻭﻟﻴﺔ ﻛﺼﻨﺪﻭﻕ ﺍﻟﻨﻘﺪ ﺍﻟﺪﻭﱄ ﻭﺍﻟﺒﻨﻚ ﺍﻟﻌﺎﳌﻲ ﻭﺍﻟﺴﻮﻕ ﺍﻷﻭﺭﺑﻴﺔ‬
‫ﺍﳌﺸﺘﺮﻛﺔ ﺍﻟﱵ ﺻﻤﻤﺖ ﳕﺎﺫﺟﻬﺎ ﺍﳋﺎﺻﺔ ‪.‬‬
‫‪6‬‬

‫ﺇﻥ ﺍﻹﺟﺮﺍﺋﲔ ﺍﳌﺬﻛﻮﺭﻳﻦ ﺃﻋﻼﻩ‪ ،‬ﻗﺪ ﻻ ﻳﺘﺮﻛﺎﻥ ﳎﺎﻻﹰ ﻛﺒﲑﺍﹰ ﻟﻠﺘﺮﺩﺩ ﰲ‬

‫ﻗﺮﺍﺭﺍﺕ ﺍﻟﺪﻭﻝ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻋﻠﻰ ﺍﳌﺴﺘﻮﻯ ﺍﻟﻜﻠﻲ ﺃﻭ ﺍﳉﺰﺋﻲ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﺆﺳﺴﺎﺕ‪،‬‬
‫ﻭﺫﻟﻚ ﻟﻠﻤﺴﺎﳘﺔ ﰲ ﺗﺒﺴﻴﻂ ﻭﺗﺮﺷﻴﺪ ﺍﻟﻘﺮﺍﺭ ﺍﻻﻗﺘﺼﺎﺩﻱ‪.‬‬

‫) ‪(Time series models‬‬ ‫‪ .2-2‬ﳕﺎﺫﺝ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‪:‬‬

‫ﳜﺘﻠﻒ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﻋﻦ ﺳﺎﺑﻘﻪ ﻣﻦ ﺣﻴﺚ ﺍﻟﺒﻨﻴﺔ ﻭﺍﳍﺪﻑ‪ ،‬ﻛﻮﻥ ﻫﺬﻩ‬
‫ﺍﻟﻨﻤﺎﺫﺝ ﺗﻘﻮﻡ ﺑﺘﻔﺴﲑ ﺍﳌﺘﻐﲑ ﺍﻟﺘﺎﺑﻊ ﺑﻮﺍﺳﻄﺔ ﺍﻟﺰﻣﻦ ﺃﻭ ﻋﱪ ﺍﻟﺴﻠﻮﻙ ﺍﳌﺎﺿﻲ ﻟﺬﻟﻚ‬
‫ﺍﳌﺘﻐﲑ‪.‬ﻓﺈﺫﺍ ﻛﺎﻧﺖ ‪ V t‬ﲤﺜﻞ ﺣﺠﻢ ﻣﺒﻴﻌﺎﺕ ﺳﻠﻌﺔ ﻣﻌﻴﻨﺔ‪ ،‬ﻓﺈﻧﻨﺎ ﻻ ﻧﺴﺘﻄﻴﻊ ﺑﺎﻻﻋﺘﻤﺎﺩ ﻋﻠﻰ‬
‫ﺍﻟﻨﻈﺮﻳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻣﻌﺮﻓﺔ ﺃﺳﺒﺎﺏ ﺍﻟﺘﻐﲑﺍﺕ ﺍﳊﺎﺻﻠﺔ ﰲ ﺣﺠﻢ ﺍﳌﺒﻴﻌﺎﺕ ﺑﺪﻗﹼﺔ‪،‬‬

‫‪6 - Les modèles appliqués de la macroéconomie, Anne Epaulard, Dunod, p18.‬‬

‫‪19‬‬
‫ﻓﻴﻤﻜﻦ ﺃﻥ ﺗﻜﻮﻥ ﻫﺬﻩ ﺍﻟﺘﻘﻠﺒﺎﺕ ﺍﺳﺘﺠﺎﺑﺔ ﻟﺘﻐﻴ‪‬ﺮ ﺍﻷﺳﻌﺎﺭ‪ ،‬ﺍﻟﺪﺧﻞ ﺍﳌﺘﺎﺡ‪... ،‬ﺍﱁ‪،‬‬
‫ﻛﻤﺎ ﺃﻧﻪ ﳝﻜﻦ ﺃﻥ ﻳﻜﻮﻥ ﻧﺎﺗﺞ ﻋﻦ ﺗﺄﺛﲑ ﻋﻮﺍﻣﻞ ﻣﻮﺿﻮﻋﻴﺔ ﺃﺧﺮﻱ ﻻ ﻧﺴﺘﻄﻴﻊ‬
‫ﻗﻴﺎﺳـﻬﺎ ﻛﺎﻟﻄـﻘﺲ‪ ،‬ﺗﻐﻴـﺮ ﺫﻭﻕ ﺍﳌﺴﺘﻬﻠﻜﲔ‪ ،‬ﻳﻮﻡ ﻣﻌﲔ ﺃﻭ ﻋﻴﺪ‪...‬ﺍﱁ‪ .‬ﺇﺫﺍ‬
‫ﳝﻜﻦ ﺗﻔﺴﲑ ﻫﺬﻩ ﺍﳌﺒﻴﻌﺎﺕ ﺑـ‪:‬‬
‫ﺍ‪ -‬ﺍﻟﺰﻣﻦ‪ :‬ﻣﻦ ﺧﻼﻝ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫) ‪V t= f (t, e t‬‬ ‫)‪(1.5‬‬

‫ﺣﻴﺚ ‪ e t‬ﲤﺜﻞ ﺍﳋﻄﺄ ﺍﻟﻌﺸﻮﺍﺋﻲ )‪ (white noise‬ﺍﳌﻌﱪ ﻋﻦ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﱵ ﻻ‬

‫ﳝﻜﻦ ﻗﻴﺎﺳﻬﺎ ﻭﻛﺬﺍ ﺍﻷﺧﻄﺎﺀ ﺍﻟﻮﺍﺭﺩﺓ ﺃﺛﻨﺎﺀ ﻋﻤﻠﻴﺔ ﲨﻊ ﻭﺇﺣﺼﺎﺀ ﻓﺘﺪﻭﻳﻦ ﺍﳌﻌﻠﻮﻣﺎﺕ‬
‫ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﺳﺎﺑﻘﺎﹰ‪ ،‬ﺣﻴﺚ ﺍﻋﺘﻤﺪ ﺗﻐﻴﲑ ﺗﺴﻤﻴﺘﻬﺎ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ ﺇﱃ ﳕﺎﺫﺝ‬
‫ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻫﺬﺍ ﺍﻟﺘﻐﻴﲑ ﺷﻜﻠﻲ ﻟﻴﺲ ﻟﻪ ﺃﻱ ﺗﺄﺛﲑ ﻣﻨﻬﺠﻲ ﻋﻠﻰ ﺍﳌﻮﺿﻮﻉ‪.‬‬
‫ﻳﺘﻢ ﺗﻨﺎﻭﻝ ﻫﺬﺍ ﺍﳌﻮﺿﻮﻉ ﺑﺸﻜﻞ ﻣﻦ ﺍﻟﺘﻔﺼﻴﻞ ﻭﺍﻟﺘﺪﻗﻴﻖ ﺿﻤﻦ ﻓﺼﻮﻝ ﺍﻟﺒﺎﺏ‬
‫ﺍﻷﻭﻝ‪.‬‬
‫ﺑـ‪ -‬ﺍﻟﺴﻠﻮﻙ ﺍﳌﺎﺿﻲ ﻟﻠﻤﺘﻐﲑ‪ :‬ﲟﻌﲎ ﺃﻧﻪ ﻳﺘﻢ ﺗﻔﺴﲑ ﺍﳌﺘﻐﲑ ﻗﻴﺪ ﺍﻟﺪﺭﺍﺳﺔ‬
‫ﺑﺴﻠﻮﻛﻪ ﺍﳌﺎﺿﻲ ﺍﳌﺘﻤﺜﻞ ﰲ ﺍﺳﺘﻌﻤﺎﻝ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ ﻭﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‬
‫ﻭﻓﻖ ﻃﺮﻳﻘﺔ ﺑﻮﻛﺲ‪ -‬ﺟﺎﻧﻜﻴﻨﺰ )‪ .(Box-Jenkins‬ﳝﻜﻦ ﲤﺜﻴﻠﻬﺎ ﰲ ﺷﻜﻠﻬﺎ ﺍﻟﺒﺴﻴﻂ‬
‫ﺍﻟﺘﺎﱄ‪:‬‬
‫) ‪Vt = f (Vt -1,Vt -2 ,...,e t‬‬ ‫)‪(1.6‬‬

‫ﺣﻴﺚ ‪ V‬ﻭ ‪ V‬ﲤﺜﻞ ﺍﳌﺒﻴﻌﺎﺕ ﰲ ﺍﻟﻔﺘﺮﺓ )‪ (t‬ﻭﺍﻟﻔﺘﺮﺓ ﺍﻟﱵ ﻗﺒﻠﻬﺎ)‪ (t-1‬ﻫﻜﺬﺍ‬

‫‪t -1‬‬ ‫‪t‬‬

‫ﺣﺴﺐ ﺩﺭﺟﺔ ﺍﻟﺘﺄﺧﲑ ﺍﳌﺮﻏﻮﺑﺔ ﺍﻟﱵ ﻻ ﲢﺪﺩ ﻋﺸﻮﺍﺋﻴﺎﹰ ﺇﳕﺎ ﺇﺣﺼﺎﺋﻴﺎ ﺑﺎﺳﺘﺨﺪﺍﻡ‬
‫ﺍﺧﺘﺒﺎﺭﺍﺕ ﻣﻨﺎﺳﺒﺔ ﻳﺄﰐ ﺫﻛﺮﻫﺎ ﰒ ﺍﻟﺘﻄﺮﻕ ﺇﻟﻴﻬﺎ ﰲ ﺍﻟﻔﺼﻠﲔ ﺍﻷﺧﲑﻳﻦ ﻣﻦ ﻫﺬﺍ‬
‫ﺍﻟﻜﺘﺎﺏ‪.‬‬
‫ﻳ‪‬ﻠﺠﺄ ﺇﱃ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﰲ ﺣﺎﻟﺔ ﻏﻴﺎﺏ ﺍﻟﻌﻼﻗﺎﺕ ﺍﻟﺴﺒﺒﻴﺔ ﺑﲔ‬
‫ﺍﳌﺘﻐﲑﺍﺕ ﺃﻭ ﻋﺪﻡ ﺗﻮﻓﺮ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﻜﺎﻓﻴﺔ ﺣﻮﻝ ﺍﳌﺘﻐﲑﺍﺕ ﺍﳌﺴﺘﻘﻠﺔ ﻭﺍﻟﻮﺍﻗﻌﺔ ﻋﻠﻰ‬

‫‪20‬‬
‫ﳝﲔ ﺍﳌﻌﺎﺩﻟﺔ‪ .‬ﻫﺬﺍ ﻻ ﻳﻌﲏ ﺃﻥ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﻏﲑ ﻣﺮﻏﻮﺏ ﻓﻴﻬﺎ ﰲ ﺍﳊﺎﻻﺕ‬
‫ﺍﻷﺧﺮﻯ‪ ،‬ﻓﺒﺴﺒﺐ ﺿﻌﻒ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ ﻋﻠﻰ ﺍﻟﺼﻌﻴﺪﻳﻦ ﺍﻹﺣﺼﺎﺋﻲ ﻭﺍﻟﺘﻨﺒﺆﻱ‬
‫ﻣﻘﺎﺭﻧﺔﹰ ﺑﺎﻹﻣﻜﺎﻧﻴﺎﺕ ﺍﳌﺴﺘﻌﻤﻠﺔ‪ ،‬ﺯﺍﺩ ﻣﻦ ﺍﻟﺮﻏﺒﺔ ﰲ ﺍﺳﺘﻌﻤﺎﳍﺎ‪ ،‬ﻛﻮ‪‬ﺎ ﻻ ﲢﺘﺎﺝ ﺇﱃ‬
‫ﳎﻬﻮﺩ ﻛﺒﲑ ﰲ ﲨﻊ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﳋﺎﺻﺔ ﺑﺎﻟﻈﺎﻫﺮﺓ ﻣﻮﺿﻮﻉ ﺍﻟﺪﺭﺍﺳﺔ‪.‬‬

‫‪21‬‬
22
 ‫اﻟﺒﺎب اﻷول‬
‫ﳕﺎذج اﻻﺳﺘﻘﻄﺎب‬
 ‫ﺍﻟﻔﺼﻞ ﺍﻟﺜﺎﱐ‬

‫ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻭﻃﺮﻕ ﻛﺸﻔﻬﺎ‬

‫ﻗﺒﻞ ﺍﻟﻮ‪‬ﻟﻮﺝ ﰲ ﻫﺬﺍ ﺍﳌﻮﺿﻮﻉ‪ ،‬ﻧﺬﻛﺮ ﻣﺮﺓ ﺃﺧﺮﻯ ﺑﺒﻌﺾ ﺩﻭﺍﻋﻲ ﺍﺳﺘﻌﻤﺎﻝ‬
‫ﳕﺎﺫﺝ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﰲ ﻋﻤﻠﻴﺎﺕ ﺍﻟﺘﻨﺒﺆ ﺍﻟﻘﺼﲑ ﺍﳌﺪﻯ ﻭﻫﻲ‪:‬‬
‫ﺍ‪ -‬ﻏﻴﺎﺏ ﺍﻟﻌﻼﻗﺎﺕ ﺍﻟﺴﺒﺒﻴﺔ ﺑﲔ ﺍﳌﺘﻐﲑﺍﺕ ﻭﺻﻌﻮﺑﺔ ﻗﻴﺎﺳﻬﺎ‪،‬‬
‫‪Explanatory‬‬ ‫ﺑـ‪ -‬ﻋﺪﻡ ﺗﻮﻓﺮ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﻜﺎﻓﻴﺔ ﺣﻮﻝ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﺸﺎﺭﺣﺔ‬
‫)‪،variables‬‬
‫ﺟـ‪ -‬ﰲ ﺣﺎﻟﺔ ﺭﻓﺾ ﳕﻮﺫﺝ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ ﺍﻗﺘﺼﺎﺩﻳﺎﹰ‪،‬ﺇﺣﺼﺎﺋﻴﺎ ﻭﺗﻨﺒﺆﻳﺎ‬
‫ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﻟﻮﺳﺎﺋﻞ ﺍﻹﺣﺼﺎﺋﻴﺔ ﺍﳌﻨﺎﺳﺒﺔ‪ .‬ﻳﺘﻢ ﺗﻨﺎﻭﻝ ﻫﺬﺍ ﺍﳌﻮﺿﻮﻉ ﺿﻤﻦ ﺍﻟﻔﺼﻞ‬
‫ﺍﻟﺜﺎﻟﺚ ﺃﺛﻨﺎﺀ ﺩﺭﺍﺳﺔ ﺗﻘﻴﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ )‪ (Statistical validation‬ﻗﺒﻞ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ‪،‬‬
‫ﻭﰲ ﺍﻟﻔﺼﻞ ﺍﻟﺴﺎﺩﺱ ﺿﻤﻦ ﻣﻮﺿﻮﻉ ﻗﻴﺎﺱ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ‪،‬‬
‫ﺩ‪ -‬ﺑﺴﺎﻃﺔ ﺗﺮﻛﻴﺐ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﻭﺳﻬﻮﻟﺔ ﺗﻔﺴﲑ ﻧﺘﺎﺋﺠﻬﺎ ﺑﺎﻟﻨﺴﺒﺔ‬
‫ﻟﻠﻤﺴﲑﻳﻦ‪،‬‬
‫ﻫـ‪ -‬ﺇﺿﺎﻓﺔ ﺇﱃ ﻛﻞ ﻫﺬﺍ‪ ،‬ﻓﺈﻥ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ ﻭﺭﻏﻢ ﺍﺳﺘﻌﻤﺎﳍﺎ‬
‫ﳌﻌﻠﻮﻣﺎﺕ ﻣﻌﺘﱪﺓ ﻭﺗﻄﻠﺒﻬﺎ ﻹﻣﻜﺎﻧﺎﺕ ﻋﻠﻤﻴﺔ ﻭﺑﺸﺮﻳﺔ ﻛﺒﲑﺓ‪ ،‬ﻓﺈﻥ ﻧﺘﺎﺋﺠﻬﺎ ﻟﻴﺴﺖ‬
‫ﺩﻭﻣﺎﹰ ﰲ ﻣﺴﺘﻮﻯ ﻫﺬﺍ ﺍ‪‬ﻬﻮﺩ ‪.‬‬
‫‪1‬‬

‫‪ .1‬ﺍﳌﺮﻛﺒﺎﺕ ﺍﳉﻮﻫﺮﻳﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬

‫ﻳﻘﺼﺪ ‪‬ﺎ ﺍﻟﻌﻨﺎﺻﺮ ﺍﳌﻜﻮﻧﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﻫﻲ ﺗﻔﻴﺪ ﰲ ﲢﺪﻳﺪ ﺳﻠﻮﻛﻬﺎ‬
‫ﰲ ﺍﳌﺎﺿﻲ ﻭﻛﺬﺍ ﺍﳌﺴﺘﻘﺒﻞ‪ ،‬ﻭﳝﻜﻦ ﺇﺩﺭﺍﺝ ﻫﺬﻩ ﺍﳌﺮﻛﺒﺎﺕ ﰲ ﺍﻟﻌﻨﺎﺻﺮ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫ﺫﻟﻚ ‪Evaluation of Econometric Models, J. Kmenta, B. Ramsey,p339.‬‬ ‫‪ - 1‬ﺃﻧﻈﺮ ﰲ‬

‫‪25‬‬
‫‪ .1-1‬ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ )‪ :(The trend‬ﺗﻌﱪ ﻋﻦ ﺍﻟﺘﻄﻮﺭ ﲟﻴﻞ ﻣﻮﺟﺐ ﺃﻭ ﺳﺎﻟﺐ‬
‫ﳌﺘﻐﲑ ﺍﻗﺘﺼﺎﺩﻱ ﻋﱪ ﺍﻟﺰﻣﻦ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﻓﻬﻲ ﺗﻌﻜﺲ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻟﻠﻈﺎﻫﺮﺓ ﺍﳌﺪﺭﻭﺳﺔ‪.‬‬
‫ﻭﻳﺮﻣﺰ ﻟﻪ ﻻﺗﻴﻨﻴﺎﹼ ﺑﺎﳊﺮﻑ‪.L :‬‬
‫‪ 1980‬ﺇﱃ ‪.2007‬‬ ‫اﻟﺸﻜﻞ )‪ :(2‬ﺍﻟﺪﺧﻞ ﺍﻟﻮﻃﲏ ﺍﳉﺰﺍﺋﺮﻱ ﻣﻦ‬

‫‪World tables, world bank, .2008‬‬ ‫ﺍﳌﺼﺪﺭ‪:‬‬

‫ﺇﻥ ﻣﻨﺤﲎ ﺍﻟﺪﺧﻞ ﺍﻟﻮﻃﲏ ﺍﳉﺰﺍﺋﺮﻱ ﻳﻌﻜﺲ ﺑﻮﺿﻮﺡ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﱵ ﻣﺎ‬
‫ﻓﺘﺌﺖ ﺗﺰﺩﺍﺩ ﻋﱪ ﺍﻟﺰﻣﻦ ﻭﲟﻴﻞ ﻣﻮﺟﺐ‪.‬‬

‫‪ .2-1‬ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ )ﺍﳌﻮﲰﻴﺔ( )‪ :(Seasonality‬ﻫﻲ ﻋﺒﺎﺭﺓ ﻋﻦ ﺗﺄﺛﲑﺍﺕ ﺧﺎﺭﺟﻴﺔ‬

‫ﺗﻄﺮﺃ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺑﻄﺮﻳﻘﺔ ﻣﻨﺘﻈﻤﺔ‪ .‬ﳝﻜﻦ ﺗﺴﻤﻴﺘﻬﺎ ﺑﺎﳌﺮﻛﺒﺔ ﺍﻟﺪﻭﺭﻳﺔ‪،‬‬
‫ﻭﻳﺮﻣﺰ ﳍﺎ ﺏ‪ .S :‬ﺗﻌﺘﱪ ﺍﻟﻔﺼﻠﻴﺔ ﻣﻦ ﺃﺑﺴﻂ ﺍﳌﺮﻛﺒﺎﺕ ﰲ ﳎﻤﻮﻋﺘﻬﺎ ﻣﻦ ﺣﻴﺚ‬
‫ﺍﻟﺘﻌﺎﻣﻞ ﻭﲢﺪﻳﺪ ﻓﺘﺮﺍﺕ ﻭﻗﻮﻋﻬﺎ‪ .‬ﻛﻤﺎ ﺃﻥ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻮﻗﻊ ﰲ ﻫﺬﻩ ﺍﳊﺎﻻﺕ ﺗﺮﺗﻜﺰ‬
‫ﻋﻠﻰ ﺣﺠﻢ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ﻭﻟﻴﺲ ﻋﻠﻰ ﻓﺘﺮﺓ ﺍﳊﺪﻭﺙ‪ .‬ﺇﻥ ﻟﻠﻤﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺩﻭﺭﻳﺔ‬
‫ﻛﺄﻥ ﲤﺜﻞ ﺍﻟﺸﻬﻮﺭ ﰲ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺸﻬﺮﻳﺔ ﺃﻭ ﺍﳌﻮﺳﻢ ﰲ ﺍﻟﺒﻴﺎﻧﺎﺕ ﺍﻟﻔﺼﻠﻴﺔ ﻭﺍﻟﱵ‬
‫ﻧﺮﻣﺰ ﳍﺎ ﺿﻤﻦ ﻫﺬﺍ ﺍﻟﺒﺎﺏ ﺑﺎﳊﺮﻑ ﺍﻟﻼﺗﻴﲏ )‪. (P‬‬

‫‪26‬‬
 ‫ﺍﻟﺸﻜﻞ )‪ :(3‬ﺍﻻﺳﺘﻬﻼﻙ ﺍﻹﲨﺎﱄ ﻟﱪﻳﻄﺎﻧﻴﺎ‬

‫ﺍﳌﺼﺪﺭ‪Economic Trends, Annual supplement U.K 1983 :‬‬

‫ﻣﻦ ﺍﳌﻼﺣﻆ ﰲ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ‪ ،‬ﺃﻥ ﻫﺬﻩ ﺍﻟﻈﺎﻫﺮﺓ ﲢﺪﺙ ﺧﻼﻝ ﺍﻟﺴﻨﺔ‬
‫ﻭﻋﻨﺪ ﺍﺳﺘﻌﻤﺎﻝ ﻣﻌﻄﻴﺎﺕ ﻓﺼﻠﻴﺔ‪ ،‬ﺷﻬﺮﻳﺔ ﺃﻭ ﺃﺳﺒﻮﻋﻴﺔ‪...‬ﺍﱁ‪ ،‬ﻭﺧﲑ ﻣﺜﺎﻝ ﻋﻠﻲ‬
‫ﺫﻟﻚ‪ ،‬ﺍﺳﺘﻬﻼﻙ ﺍﳌﺸﺮﻭﺑﺎﺕ ﺍﻟﺒﺎﺭﺩﺓ ﰲ ﺍﳉﺰﺍﺋـﺮ‪ ،‬ﺣﻴﺚ ﻳﺰﺩﺍﺩ ﺍﻟﻄﻠﺐ ﻋﻠﻴﻬﺎ ﰲ‬
‫ﻓﺼﻞ ﺍﻟﺼﻴﻒ ﻭﻳﻨﻘﺺ ﰲ ﺍﻟﺸﺘﺎﺀ‪ ،‬ﻭﻧﻔﺲ ﺍﻟﻈﺎﻫﺮﺓ ﺗﺴﺠﻞ ﰲ ﺍﻟﻄﻠﺐ ﻋﻠﻰ‬
‫ﺍﻟﻜﻬﺮﺑﺎﺀ ﰲ ﺣﲔ ﺃﻥ ﺍﻟﻄﻠﺐ ﻋﻠﻰ ﺍﺳﺘﻬﻼﻙ ﺍﻟﻐﺎﺯ ﺍﻟﻄﺒﻴﻌﻲ ﻳﺰﺩﺍﺩ ﰲ ﺍﻟﻔﺼﻮﻝ‬
‫ﺍﻟﺒﺎﺭﺩﺓ ﻭﻳﻨﻘﺺ ﰲ ﺗﻠﻚ ﺍﳊﺎﺭﺓ‪.‬‬
‫اﻟﺸﻜﻞ)‪ :(4‬ﻋﺪﺩ ﻋﻘﻮﺩ ﺍﻟﺰﻭﺍﺝ ﺣﺴﺐ ﺍﻟﻔﺼﻞ‬

‫‪.2008.4 - 1994.1‬‬ ‫ﺍﳌﺼﺪﺭ‪ :‬ﺑﻠﺪﻳﺔ ﻣﺮﺍﺩ‪ ،‬ﻭﻻﻳﺔ ﺗﻴﺒﺎﺯﺓ‪ ،‬ﺍﳊﺎﻟﺔ ﺍﳌﺪﻧﻴﺔ‬

‫‪27‬‬
‫ﻣﻦ ﺍﻟﻈﻮﺍﻫﺮ ﺍﻻﺟﺘﻤﺎﻋﻴﺔ ﺍﻟﻌﺎﻛﺴﺔ ﻟﻠﻤﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺑﺸﻜﻞ ﺟﻠﻲ‪ ،‬ﻗﻀﻴﺔ‬
‫ﺃﻓﺮﺍﺡ ﺍﻟﺰﻭﺍﺝ ﻛﻤﺎ ﻫﻮ ﻣﺒﲔ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺴﺎﺑﻖ‪:‬‬
‫‪ .3-1‬ﻣﺮﻛﺒﺔ ﺍﻟﺪﻭﺭﺍﺕ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ )‪ :(Business cycles‬ﺗﻨﻌﻜﺲ ﻫﺬﻩ ﺍﳌﺮﻛﺒﺔ ﰲ‬
‫ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﻄﻮﻳﻠﺔ ﺍﻷﺟﻞ‪ ،‬ﻭﺍﻟﱵ ﺗ‪‬ﱪﺯ ﺃﺛﺮ ﺍﻧﺘﻘﺎﻝ ﺍﻷﺣﻮﺍﻝ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻣﺜﻼ‪،‬‬
‫ﻣﻦ ﺍﻟﻜﺴﺎﺩ ﺇﱃ ﺍﻻﻧﺘﻌﺎﺵ ﻓﺎﻟﺮﻭﺍﺝ ﰒ ﺍﻟﺮﻛﻮﺩ ﻭﻫﻜﺬﺍ ﺩﻭﺍﻟﻴﻚ‪ ،‬ﻭﻫﻲ ﲤﺜﻞ ﻛﺬﻟﻚ‬
‫ﺗﺄﺛﲑ ﻋﻮﺍﻣﻞ ﺧﺎﺭﺟﻴﺔ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺑﺸﻜﻞ ﻣﻨﺘﻈﻢ ﻛﻤﺎ ﻳﻈﻬﺮ ﰲ ﺍﻟﺸﻜﻞ‬
‫)‪ ،(5‬ﺣﻴﺚ ﻧﺴﺠﻞ ﻣﺜﻼﹰ ﻓﺘﺮﰐ ﺭﻛﻮﺩ ﻣﻦ ‪ 1984)1984.1‬ﺍﻟﻔﺼﻞ ﺍﻷﻭﻝ( ﺇﱃ ﻏﺎﻳﺔ‬
‫‪ 1986) 1986.3‬ﺍﻟﻔﺼﻞ ﺍﻟﺜﺎﻟﺚ( ﺑﺎﻟﻨﺴﺒﺔ ﻟﻸﻭﱃ ﻭﻣﻦ ‪ 1993.1-1992.2‬ﺑﺎﻟﻨﺴﺒﺔ‬
‫ﻟﻔﺘﺮﺓ ﺍﻟﺮﻛﻮﺩ ﺍﻟﺜﺎﻧﻴﺔ‪ ،‬ﻭﻫﺬﻩ ﺍﻟﻈﺎﻫﺮﺓ ﲣﺘﻠﻒ ﻋﻦ ﺍﻟﺪﻭﺭﻳﺔ )ﺍﻟﻔﺼﻠﻴﺔ( ﰲ ﺃ‪‬ﺎ ﲢﺪﺙ‬
‫ﰲ ﳎﺎﻝ ﺃﻛﱪ ﻣﻦ ﺳﻨﺔ‪ ،‬ﰲ ﺣﺮﻛﺔ ﺩﻭﺭﻳﺔ ﺗﻈﻬﺮ ﰲ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﻄﻮﻳﻠﺔ‪ .‬ﻳﺮﻣﺰ‬
‫ﳍﺬﻩ ﺍﳌﺮﻛﺒﺔ ﺏ‪.C:‬‬
‫اﻟﺸﻜﻞ )‪ :(5‬ﺳﻌﺮ ﻛﺮﺍﺀ ﺳﻔﻦ ﺷﺤﻦ ﺍﻟﻄﺎﻗﺔ )‪ 9-5‬ﺁﻻﻑ ﻃﻦ( ﰲ ﺍﻟﺴﻮﻕ ﺍﻟﺪﻭﱄ‬
‫)‪(1994.4- 1981.1‬‬

‫ﺍﳌﺼﺪﺭ‪ :‬ﺳﻮﻧﺎﻃﺮﺍﻙ‪ ،‬ﺍﻟﻘﺴﻢ ﺍﻟﺘﺠﺎﺭﻱ‪.‬‬

‫‪ .4-1‬ﺍﳌﺮﻛﺒﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ )‪ :(Irregular component‬ﺗﻌﱪ ﻋﻦ ﺍﻟﺬﺑﺬﺑﺎﺕ ﻏﲑ‬
‫ﺍﳌﻨﺘﻈﻤﺔ ﻭﺍﳌﻌﱪﺓ ﻋﻦ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﱵ ﺗﻨﺘﻬﺞ ﺳﻠﻮﻛﺎﹰ ﻋﺸﻮﺍﺋﻴﺎﹰ‪ ،‬ﺗﺘﺬﺑﺬﺏ ﻫﺬﻩ ﺍﳌﺮﻛﺒﺔ‬
‫ﺣﻮﻝ ﻭﺳﻂ ﺣﺴﺎﰊ ﺛﺎﺑﺖ ﻧﺴﺒﻴﺎﹰ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ﺍﻟﺰﻣﻨﻴﺔ ﺍﳌﺪﺭﻭﺳﺔ‪ .‬ﺗﱪﺯ ﻫﺬﻩ‬

‫‪28‬‬
‫ﺍﻟﻈﺎﻫﺮﺓ ﳌﺎ ﻳﻜﻮﻥ ﺳﻠﻮﻙ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻣﺴﺘﻘﺮﺍﹰ ﻭﻣﻨﻪ ﻧﺘﻮﻗﻊ ﺃﻥ ﺗﻜﻮﻥ ﺗﻨﺒﺆﺍ‪‬ﺎ‬
‫ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﺛﺎﺑﺘﺔ ﻭﻣﻌﺎﺩﻟﺔ ﻟﻮﺳﻄﻬﺎ ﺍﳊﺴﺎﰊ‪ .‬ﻳـﺮﻣﺰ ﳍﺬﻩ ﺍﳌﺮﻛﺒﺔ ﺏ‪.I :‬‬
‫)‪(2006- 1963‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(6‬ﺇﻧﺘﺎﺝ ﺍﳊﺒﻮﺏ ﰲ ﺍﳉﺰﺍﺋﺮ‬

‫ﺍﳌﺼﺪﺭ‪ :‬ﻭﺯﺍﺭﺓ ﺍﻟﻔﻼﺣﺔ ﻭﺍﻟﺘﻨﻤﻴﺔ ﺍﻟﺮﻳﻔﻴﺔ‪.‬‬

‫‪ .2‬ﻛﺸﻒ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬

‫ﺑﺎﻟﻨﻈﺮ ﺇﱃ ﺍﻟﺸﻜﻞ )‪ ،(7‬ﳝﻜﻦ ﺍﻟﺘﺤﻘﻖ ﰲ ﺑﻌﺾ ﺍﳊﺎﻻﺕ ﻣﻦ ﺻﻌﻮﺑﺔ‬
‫ﻛﺸﻒ ﻣﺮﻛﺒﺔ ﺃﻭ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻗﻴﺪ ﺍﻟﺪﺭﺍﺳﺔ‪ ،‬ﻭﳍﺬﺍ ﺳﻨﺘﻨﺎﻭﻝ ﺑﻌﺾ‬
‫ﺍﻟﻄﺮﻕ ﺍﳌﺴﺎﻋﺪﺓ ﻋﻠﻰ ﻫﺬﺍ ﺍﻟﻜﺸﻒ‪.‬‬

‫‪Plot‬‬ ‫‪ .1-2‬ﺍﻟﺘﺤﻠﻴﻞ ﺍﻟﺒﻴﺎﱐ ﻟﻠﻤﻌﻄﻴﺎﺕ‬

‫‪‬ﺘﻢ ﰲ ﻫﺬﻩ ﺍﳌﺮﺣﻠﺔ ﺑﺪﺭﺍﺳﺔ ﻭﲢﻠﻴﻞ ﺍﻟﻈﺮﻭﻑ ﺍﻟﱵ ﺗﻮﻟﺪﺕ ﻋﻨﻬﺎ ﻫﺬﻩ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻓﺈﺫﺍ ﻛﺎﻥ ﻫﺬﺍ ﺍﶈﻴﻂ ﻣﺴﺘﻘﺮﺍﹰ ﺗﻜﻮﻥ ﺍﻟﺴﻠﺴﻠﺔ ﻛﺬﻟﻚ‪ ،‬ﻭﺍﻟﻌﻜﺲ‬
‫ﺻﺤﻴﺢ‪ .‬ﺗﺘﻤﺔ ﳍﺬﺍ ﺍﻟﻌﻤﻞ ﻧﻘﻮﻡ ﺑﺘﻤﺜﻴﻞ ﻫﺬﻩ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﻟﺮﻗﻤﻴﺔ ﰲ ﺷﻜﻞ ﺑﻴﺎﱐ‪،‬‬
‫ﻳﻌﻜﺲ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺑﺸﻜﻞ ﺃﻭﺿﺢ‪ .‬ﻓﻴﺘﻤﺜﻞ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺗﻠﻚ‬
‫ﺍﳌﺮﻛﺒﺔ ﺍﻟﱵ ﺗﺪﻓﻊ ﺑﺎﳌﻨﺤﲎ ﳓﻮ ﺍﻟﺰﻳﺎﺩﺓ‪ ،‬ﺇﺫﺍ ﻛﺎﻥ ﻣﻴﻠﻬﺎ ﻣﻮﺟﺒ‪‬ﺎ‪ ،‬ﺃﻭ ﺇﱃ ﺍﻟﻨﻘﺼﺎﻥ ﺇﺫﺍ‬
‫ﻛﺎﻥ ﺳﺎﻟﺒﺎﹰ‪ .‬ﺑﻴﻨﻤﺎ ﺗﻨﻌﻜﺲ ﺍﳌﺮﻛﺒﺔ ﺍﻟﺪﻭﺭﻳﺔ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺒﻴﺎﱐ ﻋﻠﻰ ﻫﻴﺌﺔ ﻗﻤﻢ‬
‫)‪ (Peaks‬ﺃﻭ ﺍﳔﻔﺎﺿﺎﺕ )ﻧﺘﻮﺁﺕ( )‪ (Troughs‬ﺑﺸﻜﻞ ﻣﻨﺘﻈﻢ‪ ،‬ﻳﺴﻤﺢ ﻟﻨﺎ ﺑﺘﺤﺪﻳﺪ‬
‫ﻓﺘﺮﺓ ﺣﺪﻭﺙ ﻫﺬﻩ ﺍﻟﻈﺎﻫﺮﺓ‪ ،‬ﻛﺄﻥ ﺗﻜﻮﻥ ﰲ ﻓﺼﻞ ﺃﻭ ﺷﻬﺮ ﻣﻌﲔ‪...‬ﺍﱁ‪.‬‬

‫‪29‬‬
 ‫ﺍﻟﺸﻜﻞ )‪ :(7‬ﺍﺳﺘﻬﻼﻙ ﻗﺎﺭﻭﺭﺍﺕ ﻏﺎﺯ ‪13‬ﻛﻎ ﰲ ﻭﻻﻳﺎﺕ ﺍﻟﻮﺳﻂ ﺍﳋﻤﺲ ﻟﻠﺠﺰﺍﺋﺮ‬
‫)‪(.92.02- 87.01‬‬

‫ﺍﳌﺼﺪﺭ‪ :‬ﺳﻮﻧﺎﻃﺮﺍﻙ‪ ،‬ﻭﺣﺪﺓ ﺣﺴﲔ ﺩﺍﻱ‪ ،‬ﺍﳉﺰﺍﺋﺮ‪.‬‬

‫ﺑﻴﻨﻤﺎ ﺗﺘﻤﺜﻞ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﰲ ﺗﻠﻚ ﺍﻟﺬﺑﺬﺑﺎﺕ ﺍﻟﱵ ﺗﺸﻮﺵ ﻋﻠﻰ ﻳﻠﻮﻙ‬
‫ﺍﳌﺮﻛﺒﺎﺕ ﺍﳌﻨﺘﻈﻤﺔ ﻭﺗﻄﺒﻌﻬﺎ ﺑﺼﺒﻐﺔ ﻋﺸﻮﺍﺋﻴﺔ‪.‬ﻛﻤﺎ ﻳﻈﻬﺮ ﺑﻴﺎﻧﻴﺎ ﺩﺍﺋﻤﺎ‪ ،‬ﻭﺑﺎﻟﻨﻈﺮ ﺇﱃ‬
‫ﺍﻟﺸﻜﻞ )‪ (3‬ﺍﻟﻮﺍﺭﺩ ﰲ ﺍﻟﺼﻔﺤﺔ )‪ ،(14‬ﳝﻜﻦ ﺑﺴﻬﻮﻟﺔ ﻛﺸﻒ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻣﻦ‬
‫ﺧﻼﻝ ﺍﻻﻧﺘﻈﺎﻡ ﺍﳌﻮﺟﻮﺩ ﰲ ﺗﺴﺠﻴﻞ ﻗﻤﺔ ﰲ ﺍﻟﻔﺼﻞ ﺍﻷﺧﲑ ﻟﻜﻞ ﺳﻨﺔ ﻭﺍﳔﻔﺎﺽ‬
‫ﻭﺍﺿﺢ ﰲ ﺑﺪﺍﻳﺔ ﻛﻞ ﺳﻨﺔ ﺟﺪﻳﺪﺓ‪.‬‬
‫‪Statistical tests‬‬ ‫‪ .2-2‬ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﻹﺣﺼﺎﺋﻴﺔ‬
‫ﺇﻻ ﺃﻧﻪ ﰲ ﻛﺜﲑ ﻣﻦ ﺍﳊﺎﻻﺕ‪ ،‬ﻻ ﻳﻜﻮﻥ ﺍﺧﺘﺒﺎﺭ ﺍﻟﻜﺸﻒ ﺍﻟﺒﻴﺎﱐ ﻛﺎﻓﻴﹰﺎ‬
‫ﻟﻮﺣﺪﻩ ﻟﻜﺸﻒ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺑﺸﻜﻞ ﺩﻗﻴﻖ ﳑﺎ ﻳﺴﺘﻠﺰﻡ ﺍﺳﺘﻌﻤﺎﻝ‬
‫ﺃﺩﻭﺍﺕ ﺇﺣﺼﺎﺋﻴﺔ ﳍﺬﺍ ﺍﻟﻐﺮﺽ‪.‬‬
‫‪ .1-2-2‬ﻛﺸﻒ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‬
‫ﳝﻜﻦ ﺗﻘﺴﻴﻢ ﻫﺬﻩ ﺍﻷﺩﻭﺍﺕ ﺇﱃ ﳎﻤﻮﻋﺘﲔ ﻣﺘﺒﺎﻳﻨﺘﲔ ﻟﻜﺸﻒ ﺃﻭﻻﱠ ﺍﻻﲡﺎﻩ‬
‫ﺍﻟﻌﺎﻡ ﰒ ﺍﻟﻔﺼﻠﻴﺔ ﰲ ﻭﻗﺖ ﻻﺣﻖ‪ ،‬ﻭﳘﺎ‪:‬‬
‫‪Non-parametric Tests‬‬ ‫ﺍ‪ -‬ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﳊﺮﺓ‬
‫ﻧﻘﺼﺪ ‪‬ﺬﺍ‪ ،‬ﺍﺳﺘﻌﻤﺎﻝ ﺍﻷﺩﻭﺍﺕ ﺍﻻﺧﺘﺒﺎﺭﻳﺔ ﺍﻟﱵ ﻻ ﲣﻀﻊ ﺑﺎﻟﻀﺮﻭﺭﺓ ﻷﻱ‬
‫ﺗﻮﺯﻳﻊ ﺇﺣﺼﺎﺋﻲ‪ ،‬ﻓﻬﻲ ﺇﺫﺍ ﺣﺮﺓ ﺍﻟﺘﻮﺯﻳﻊ‪ ،‬ﻓﻬﻲ ﻻ ﺗﺘﻄﻠﺐ ﺃﻱ ﻓﺮﺿﻴﺔ ﺣﻮﻝ ﺍﻟﺘﻮﺯﻳﻊ‬

‫‪30‬‬
‫ﺍﻻﺣﺘﻤﺎﱄ ﻟﻠﺨﻄﺄ ‪ . et‬ﻭﳍﺬﺍ ﻭﺭﻏﻢ ﺳﻬﻮﻟﺔ ﺗﻜﻮﻳﻦ ﻭﺣﺴﺎﺏ ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﻣﻦ‬
‫ﺍﻻﺧﺘﺒﺎﺭﺍﺕ‪ ،‬ﻓﺈﻧﻪ ﻳﻌﺎﺏ ﻋﻠﻴﻬﺎ ﺿﻌﻔﻬﺎ ﰲ ﻛﺸﻒ ﺍﳌﺮﻛﺒﺎﺕ ﺍﳌﻌﻨﻴﺔ‪ .‬ﳍﺬﺍ ﻭﰲ ﻫﺬﻩ‬
‫ﺍ‪‬ﻤﻮﻋﺔ ﺑﺎﻟﺬﺍﺕ‪ ،‬ﺳﻨﺤﺎﻭﻝ ﺗﺮﺗﻴﺐ ﻫﺬﻩ ﺍﳌﻘﺎﻳﻴﺲ ﺣﺴﺐ ﺍﻷﻓﻀﻠﻴﺔ ﺇﻥ ﺃﻣﻜﻦ ﺣﱴ‬
‫ﻳﺘﺴﲎ ﻟﻠﻄﺎﻟﺐ ﺍﻟﻮﻗﻮﻑ ﻋﻠﻰ ﻫﺬﻩ ﺍﻟﻨﺘﻴﺠﺔ ﺑﻨﻔﺴﻪ‪.‬‬

‫‪Run Test‬‬ ‫ﺍ‪ -1-‬ﺍﺧﺘﺒﺎﺭ ﺍﻟﺘﻮﺍﱄ )ﺗﻌﺎﻗﺐ ﺍﻹﺷﺎﺭﺍﺕ(‬

‫ﻳﺼﻠﺢ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻟﻜﺸﻒ ﻣﺪﻯ ﻋﺸﻮﺍﺋﻴﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﳍﺬﺍ ﻳﺪﻋﻰ ﰲ‬
‫ﻏﺎﻟﺐ ﺍﳌﺮﺍﺟﻊ ﺍﻹﺣﺼﺎﺋﻴﺔ ﺑﺎﺧﺘﺒﺎﺭ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ )‪ ،(Test of Randomness‬ﻭﻫﻮ‬
‫ﻳﺴﺘﻌﻤﻞ ﰲ ﺍﻟﺘﺤﻘﻖ ﻣﻦ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ .‬ﻛﻤﺎ ﻳﺴﺘﺨﺪﻡ‬
‫ﰲ ﺣﺎﻻﺕ ﺃﺧﺮﻯ ﻻﺧﺘﺒﺎﺭ ﻋﻨﺼﺮ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﰲ ﻭﻗﻮﻉ ﺍﳊﻮﺍﺩﺙ ﺍﳌﺨﺘﻠﻔﺔ‪.‬‬
‫ﺻﻴﻐﺘﻪ‪:‬‬
‫‪H0 :‬‬ ‫ﺍﻟﺴﻠﺴﻠﺔ ﻋﺸﻮﺍﺋﻴﺔ‬
‫‪HA :‬‬ ‫ﺍﻟﺴﻠﺴﻠﺔ ﺫﺍﺕ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‬
‫‪2‬‬

‫ﺗﻜﻮﻳﻦ ﺍﻻﺧﺘﺒﺎﺭ‬
‫‪ -1‬ﺗﺮﺗﻴﺐ ﻣﺸﺎﻫﺪﺍﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺣﺴﺐ ﺍﻷﳘﻴﺔ ﺃﻱ ﻣﻦ ﺍﻷﺻﻐﺮ ﺇﱃ‬
‫ﺍﻷﻛﱪ ﰲ ﺗﺮﺗﻴﺐ ﺗﺼﺎﻋﺪﻱ‪.‬‬
‫‪ -2‬ﺣﺴﺎﺏ ﺍﻟﻮﺳﻴﻂ‪ ،‬ﻭﻫﻲ ﺍﳌﺸﺎﻫﺪﺓ ﺍﳌﻘﺎﺑﻠﺔ ﻟﻠﺮﺗﺒﺔ ‪ m‬ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﺮﺗﺒﺔ‬
‫ﺗﺮﺗﻴﺒﺎﹰ ﺗﺼﺎﻋﺪﻳﺎﹰ ﻭﺍﳌﻌﻄﺎﺓ ﺑﺎﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫ﰲ ﺣﺎﻟﺔ ﻛﻮﻥ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ‪ T‬ﻓﺮﺩﻱ‪:‬‬
‫)‪( T+1‬‬
‫= ‪m‬‬
‫‪2‬‬
‫ﰲ ﺣﺎﻟﺔ ﻛﻮﻥ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ‪ T‬ﺯﻭﺟﻲ‪:‬‬
‫‪T‬‬
‫= ‪m‬‬
‫‪2‬‬

‫‪ - 2‬ﲟﻌﲎ ﺃ‪‬ﺎ ﲢﻮﻱ ﻣﺮﻛﺒﺔ ﺍﲡﺎﻩ ﻋﺎﻡ ﺇﺿﺎﻓﺔ ﺇﱃ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ‪.‬‬

‫‪31‬‬
 ‫ﺇﺫﺍ ﻓﺎﻟﻮﺳﻴﻂ ﻫﻮ ﰲ ﺍﳊﺎﻟﺔ ﺍﻷﻭﱃ ﻭﺍﻟﺜﺎﻧﻴﺔ ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪Md = ym‬‬ ‫)‪(2.1‬‬

‫) ‪( y m + y m +1‬‬ ‫)‪(2.2‬‬
‫= ‪Md‬‬
‫‪2‬‬
‫ﺣﻴﺚ ‪ y‬ﲤﺜﻞ ﺷﻌﺎﻉ ﺍﳌﺸﺎﻫﺪﺍﺕ‪ ،‬ﻣﺮﺗﺒﺔ ﺗﺮﺗﻴﺒﺎ ﺗﺼﺎﻋﺪﻳﺎ ﺃﻭ ﺗﻨﺎﺯﻟﻴﺎ ﻭﺗﺼﺒﺢ‬
‫‪ m‬ﺩﻟﻴﻠﻬﺎ‪.‬‬
‫‪ -3‬ﺇﻋﻄﺎﺀ ﺇﺷﺎﺭﺓ ﺳﺎﻟﺒﺔ )‪ (-‬ﻟﻠﻘﻴﻢ ﺍﻷﺻﻐﺮ ﻣﻦ ‪ Md‬ﻭﻣﻮﺟﺒﺔ )‪ (+‬ﻟﻸﻛﱪ‬
‫ﻭﺍﻟﱵ ﺗﺘﺸﻜﻞ ﰲ ﳎﻤﻮﻋﺔ ﺇﺷﺎﺭﺍﺕ‪.‬‬
‫‪ -4‬ﺣﺴﺎﺏ ‪ R‬ﺍﳌﻤﺜﻞ ﻟﻌﺪﺩ ﻣﺮﺍﺕ ﺗﻮﺍﱄ ﺍﻹﺷﺎﺭﺓ ﻣﻦ ﻣﻮﺟﺐ ﺇﱃ ﺳﺎﻟﺐ‬
‫ﺃﻭ ﺍﻟﻌﻜﺲ‪.‬‬
‫ﺍﻟﻘـﺮﺍﺭ‪ :‬ﺭﻓﺾ ‪ H‬ﺇﺫﺍ ﻛﺎﻥ‪:‬‬ ‫‪0‬‬

‫ﺍ‪ -‬ﰲ ﺍﻟﻌﻴﻨﺎﺕ ﺍﻟﺼﻐﲑﺓ ‪( m £ 20 ) :‬‬

‫‪ R ³ R u‬ﺃﻭ ‪R £ R‬‬
‫‪l‬‬

‫ﺍﳌﺨﻄﻂ )‪ :(1‬ﺍﺧﺘﺒﺎﺭ ﺍﻟﺘﻮﺍﱄ‪:‬‬

‫‪28‬‬ ‫‪0‬‬
‫‪RU‬‬ ‫‪RL‬‬

‫‪H‬‬ ‫‪0‬‬ ‫ﻣﻨﻄﻘﺔ ﺭﻓﺾ‬ ‫‪H‬‬ ‫‪0‬‬ ‫ﻣﻨﻄﻘﺔ ﻗﺒﻮﻝ‬ ‫‪H‬‬ ‫‪0‬‬ ‫ﻣﻨﻄﻘﺔ ﺭﻓﺾ‬
‫ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‬ ‫ﻭﺍﻟﺪﻧﻴﺎ‪3‬‬
‫ﺣﻴﺚ ‪ R l , R u‬ﲤﺜﻞ ﺍﻟﻘﻴ‪‬ﻢ ﺍﳊﺮﺟﺔ ﺍ‪‬ﺪﻭﻟﺔ ﺍﻟﻌﻠﻴﺎ‬
‫ﻭﺍﳌﻘﺎﺑﻠﺔ ﻟﻠﺮﺗﺒﺔ ‪ . m‬ﻳﺘﻢ ﻗﺒﻮﻝ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ‪ H‬ﰲ ﺍﳊﺎﻟﺔ ﺍﻷﺧﺮﻯ‪.‬‬
‫‪0‬‬

‫ﻧﺸﲑ ﺇﱃ ﺿﺮﻭﺭﺓ ﺍﻟﺘﻔﺮﻳﻖ ﺑﲔ ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ ﳍﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻭﺗﻠﻚ ﺍﳌﺘﻌﻠﻘﺔ‬

‫ﺑﺎﺧﺘﺒﺎﺭ ﺩﺍﺭﺑﻦ ﻭﻃﺴﻦ)‪ (Durbin Watson‬ﺍﳌﺴﺘﻌﺎﻥ ﺑﻪ ﰲ ﻛﺸﻒ ﻇﺎﻫﺮﺓ‬
‫ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻸﺧﻄﺎﺀ‪ ،‬ﻭﻫﻮ ﻣﻘﻴﺎﺱ ﺃﺳﺎﺳﻲ ﰲ ﻋﻤﻠﻴﺎﺕ ﺍﻟﻨﻤﺬﺟﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ‪.‬‬

‫‪ -3‬ﺃﻧﻈﺮ ﺍﳉﺪﺍﻭﻝ ﺍﻹﺣﺼﺎﺋﻴﺔ‪.‬‬

‫‪32‬‬
 ‫‪Z > Za‬‬ ‫ﺏ‪ -‬ﰲ ﺍﻟﻌﻴﻨﺎﺕ ﺍﻟﻜﺒﲑﺓ‪(m > 20 :‬‬
‫‪2‬‬

‫ﺣﻴﺚ ‪ Z‬ﻣﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪R - mR‬‬
‫= ‪Z‬‬ ‫)‪(2.3‬‬
‫‪sR‬‬
‫ﺃﻳﻦ‪:‬‬
‫‪mR = m +1‬‬ ‫)‪(2.4‬‬

‫)‪m( m +1‬‬
‫= ‪sR‬‬ ‫)‪(2.5‬‬ ‫ﻭ‬
‫‪2m - 1‬‬

‫ﻣﻦ ﺟﺪﻭﻝ ﻣﺴﺎﺣﺎﺕ ﺍﻟﺘﻮﺯﻳﻊ ﺍﻟﻄﺒﻴﻌﻲ‬ ‫‪Za‬‬ ‫ﻳﺘﻢ ﺍﳊﺼﻮﻝ ﻋﻠﻰ ﻗﻴﻤﺔ‬
‫‪2‬‬
‫‪.a‬‬ ‫‪= 5%‬‬ ‫ﺍﳌﻌﻴﺎﺭﻱ ﺣﻴﺚ ﺗﺴﺎﻭﻱ ‪ 1.96‬ﻋﻨﺪ ﻣﺎ‬
‫ﻣﺜﺎﻝ ‪ :4‬ﺗﺄﻛﺪ ﻣﻦ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ ﺍﳋﺎﺻﺔ ﺑﺴﻜﺎﻥ‬
‫ﺍﳉﺰﺍﺋﺮ ﻣﻦ ‪ 1980‬ﺇﱃ ‪ ،2006‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ‪.‬‬
‫اﻟﺠﺪول )‪ :(2‬ﺳﻜﺎن اﻟﺠﺰاﺋﺮ ﻣﻦ )‪ ( 2007 - 1980‬اﻟﻮﺣﺪة‪ :‬ﻣﻠﯿﻮن ﻧﺴﻤﺔ‬
‫‪1988‬‬ ‫‪1987‬‬ ‫‪1986‬‬ ‫‪1985‬‬ ‫‪1984‬‬ ‫‪1983‬‬ ‫‪1982‬‬ ‫‪1981‬‬ ‫‪1980‬‬ ‫ﺍﻟﺴﻨﺔ‬
‫‪24.10‬‬ ‫‪23.40‬‬ ‫‪22.80‬‬ ‫‪22.20‬‬ ‫‪21.17‬‬ ‫‪20.51‬‬ ‫‪19.86‬‬ ‫‪19.24‬‬ ‫‪18.66‬‬ ‫ﺍﻟﻤﺸﺎﻫﺩﺓ‬
‫‪1997‬‬ ‫‪1996‬‬ ‫‪1995‬‬ ‫‪1994‬‬ ‫‪1993‬‬ ‫‪1992‬‬ ‫‪1991‬‬ ‫‪1990‬‬ ‫‪1989‬‬ ‫ﺍﻟﺴﻨﺔ‬
‫‪29.04‬‬ ‫‪28.56‬‬ ‫‪28.06‬‬ ‫‪27.49‬‬ ‫‪26.89‬‬ ‫‪26.27‬‬ ‫‪25.64‬‬ ‫‪25.02‬‬ ‫‪24.70‬‬ ‫ﺍﻟﻤﺸﺎﻫﺩﺓ‬
‫‪2006‬‬ ‫‪2005‬‬ ‫‪2004‬‬ ‫‪2003‬‬ ‫‪2002‬‬ ‫‪2001‬‬ ‫‪2000‬‬ ‫‪1999‬‬ ‫‪1998‬‬ ‫ﺍﻟﺴﻨﺔ‬
‫‪33.49‬‬ ‫‪32.90‬‬ ‫‪32.36‬‬ ‫‪31.84‬‬ ‫‪31.35‬‬ ‫‪30.87‬‬ ‫‪30.41‬‬ ‫‪29.96‬‬ ‫‪29.50‬‬ ‫ﺍﻟﻤﺸﺎﻫﺩﺓ‬
‫‪2007‬‬ ‫ﺍﻟﺴﻨﺔ‬
‫‪33.99‬‬ ‫ﺍﻟﻤﺸﺎﻫﺩﺓ‬

‫ﻧﻼﺣﻆ ﻫﻨﺎ ﺃﻧﻪ ﻻ ﻳﻮﺟﺪ ﻓﺮﻕ ﺑﲔ ﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﺰﻣﲏ ﻭﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﺘﺼﺎﻋﺪﻱ‪،‬‬

‫ﺣﻴﺚ ﺃﻧﻪ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻌﻼﻗﺎﺕ ﺍﻟﺴﺎﺑﻘﺔ ﳝﻜﻦ ﺗﻜﻮﻳﻦ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻋﻦ ﻃﺮﻳﻖ‬
‫ﲢﺪﻳﺪ ﳎﻤﻮﻋﺎﺕ ﺍﻹﺷﺎﺭﺍﺕ ﺑﺪﺍﻳﺔ ﻣﻦ ﺍﻟﻴﺴﺎﺭ ﺇﱃ ﺍﻟﻴﻤﲔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪33‬‬
 ‫‪--------------‬‬ ‫‪++++++++++++++‬‬

‫ﳍﺬﺍ ﻳﺘﺒﲔ ﺃﻥ ﺍﻹﺷﺎﺭﺍﺕ ﻗﺪ ﺷﻜﻠﺖ ﳎﻤﻮﻋﺘﲔ‪ ،‬ﺍﻷﻭﱃ ﺳﺎﻟﺒﺔ ﻭ ﺍﻟﺜﺎﻧﻴﺔ‬

‫ﻣﻮﺟﺒﺔ ﻟﻴﻜﻮﻥ‪،‬‬
‫‪m =14, Md = 27.19 , R = 2‬‬

‫ﻳﺘﻢ ﺍﺳﺘﺨﺮﺍﺝ ﻗﻴﻤﱵ ﺍﳊﺪ ﺍﻷﺩﱏ ﻭﺍﻷﻋﻠﻰ ﺍﳌﻘﺎﺑﻠﺔ ﻟﻠﺮﺗﺒﺔ )‪ (m‬ﻣﻦ ﺟﺪﻭﻝ‬
‫ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ ﺍﳋﺎﺻﺔ ‪‬ﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻭﺍﻟﻮﺍﺭﺩ ﺿﻤﻦ ﺍﳌﻼﺣﻖ ﺍﻹﺣﺼﺎﺋﻴﺔ‪ .‬ﻓﺎﻟﻘﻴﻢ‬
‫ﺍﳊﺮﺟﺔ ﺍﻟﺪﻧﻴﺎ ﻭ ﺍﻟﻘﺼﻮﻯ ﳍﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﳌﺎ ﺗﻜﻮﻥ ‪ m =14‬ﻫﻲ)‪ (9‬ﻭ )‪ (20‬ﻋﻠﻰ‬
‫ﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﱵ ﳝﻜﻦ ﻗﺮﺃ‪‬ﺎ ﻣﻦ ﺍﳉﺪﻭﻝ )‪ (3‬ﺍﻟﺘﺎﱄ ﺍﻟﺬﻱ ﻳﺒﲔ ﺟﺰﺀ ﻣﻨﻪ‪ ،‬ﺇﺫﺍ‪:‬‬
‫‪R l = 9 , R u = 20‬‬
‫اﻟﺠﺪول )‪ :(3‬ﺟﺰء ﻣﻦ ﺟﺪول اﻟﻘﯿﻢ اﻟﺤﺮﺟﺔ ل ‪R‬‬
‫‪Ru‬‬ ‫‪R‬‬ ‫‪l‬‬
‫‪m‬‬
‫‪10‬‬ ‫‪2‬‬ ‫‪5‬‬
‫‪11‬‬ ‫‪3‬‬ ‫‪6‬‬
‫‪13‬‬ ‫‪3‬‬ ‫‪7‬‬
‫‪...‬‬ ‫‪...‬‬ ‫‪...‬‬
‫‪20‬‬ ‫‪9‬‬ ‫‪14‬‬
‫…‬ ‫…‬ ‫…‬
‫‪28‬‬ ‫‪14‬‬ ‫‪20‬‬

‫ﺍﳌﺼﺪﺭ‪ :‬ﺟﺪﺍﻭﻝ ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ – ﺍﳌﻠﺤﻖ‪.‬‬

‫ﺍﻟﻘﺮﺍﺭ‪ :‬ﻛﻮﻥ ‪ ، R £ R l‬ﺃﻳﻦ ﲢﻘﻖ ﺍﻟﻮﺟﻪ ﺍﻷﻭﻝ ﻣﻦ ﻫﺬﺍ ﺍﻟﺸﺮﻁ‪ .‬ﻭﺑﺎﻟﺘﺎﱄ‬
‫ﺭﻓﺾ ‪ H 0‬ﻳﻌﲏ ﻗﺒﻮﻝ ﺍﻟﻔﺮﺿﻴﺔ ﺍﻟﺒﺪﻳﻠﺔ ﻭﻣﻨﻪ ﺛﺒﻮﺕ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫ﺍﳌﺨﻄﻂ )‪ :(2‬ﺍﺧﺘﺒﺎﺭ ﺍﻟﺘﻮﺍﱄ‬
‫‪28‬‬ ‫‪20‬‬ ‫‪9‬‬ ‫‪0‬‬
‫ﻣ‬
‫‪H‬‬ ‫‪0‬‬ ‫ﻣﻨﻄﻘﺔ ﺭﻓﺾ‬ ‫‪H‬‬ ‫‪0‬‬ ‫ﻣﻨﻄﻘﺔ ﻗﺒﻮﻝ‬ ‫‪H‬‬ ‫‪0‬‬ ‫ﻧﻄﻘﺔ ﺭﻓﺾ‬

‫‪34‬‬
‫ﳍﺬﺍ ﻧﺘﻮﻗﻊ ‪ -‬ﺣﺴﺐ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ‪ -‬ﺃﻥ ﺗﻜﻮﻥ ) ‪ ( R‬ﺇﻣﺎ ﺿﻌﻴﻔﺔ ﺃﻭ ﺷﺪﻳﺪﺓ‬
‫ﻭﻣﺮﻛﺰﺓ ﰲ ﺣﺎﻟﺔ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﺃﻭ ﻣﻌﺘﱪﺓ ﰲ ﺣﺎﻟﺔ ﻋﺪﻡ ﻭﺟﻮﺩﻫﺎ‪.‬‬
‫ﻧﺬﻛﺮ ﻣﺮﺓ ﺃﺧﺮﻯ ﺃﻥ ﻗﺒﻮﻝ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ﻻ ﻳﺘﻢ ﺇﻻ ﻋﻨﺪ ﻭﻗﻮﻉ ) ‪ ( R‬ﺑﲔ ﺍﻟﻘﻴﻤﺘﲔ‬
‫ﺍ‪‬ﺪﻭﻟﺘﲔ ﺍﻟﻘﺼﻮﻯ ﻭﺍﻟﺪﻧﻴﺎ‪ .‬ﻟﺘﻮﺿﻴﺢ ﺳﻠﻮﻙ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﰲ ﺣﺎﻟﺔ ﺗﻌﺪﺩ‬
‫ﺍﳌﺮﻛﺒﺎﺕ‪ ،‬ﻧﺪﺭﺝ ﺍﳌﺜﺎﻝ ﺍﻟﺘﺎﱄ‪:‬‬
‫)‪(4‬‬ ‫ﻣﺜﺎﻝ ‪ :5‬ﻟﺪﻳﻨﺎ ﻣﻌﻄﻴﺎﺕ ﺍﻻﺳﺘﻬﻼﻙ ﺍﳌﻮﲰﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ ﻭﺍﻟﻮﺍﺭﺩﺓ ﰲ ﺍﳉﺪﻭﻝ‬
‫ﺍﳌﺘﻌﻠﻘﺔ ﺑﺎﻻﺳﺘﻬﻼﻙ ﺍﻟﻔﺼﻠﻲ ﺍﻟﱪﻳﻄﺎﱐ‪ ،‬ﺍﳌﻄﻠﻮﺏ ﲢﺪﻳﺪ ﻋﺸﻮﺍﺋﻴﺔ ﻫﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ‪.‬‬
‫ﳊﻞ ﻫﺬﺍ ﺍﳌﺜﺎﻝ ﺍﳌﺘﻌﻠﻖ ﺑﺎﻟﺒﺤﺚ ﰲ ﻣﺪﻯ ﺍﺣﺘﻮﺍﺀ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻋﻠﻰ‬
‫ﻣﺮﻛﺒﺎﺕ ﺃﺧﺮﻯ ﺇﱃ ﺟﺎﻧﺐ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪ ،‬ﻧﻘﻮﻡ ﻋﻨﺪ ﻫﺬﺍ ﺍﳌﺴﺘﻮﻯ ﺑﺎﺧﺘﺒﺎﺭ‬
‫ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻛﻤﺎ ﺳﻴﺄﰐ‪.‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(4‬ﺗﺸﻜﻴﻞ ﺍﺧﺘﺒﺎﺭ ﺗﻮﺍﱄ ﺍﻹﺷﺎﺭﺍﺕ‬
‫ﺍﳌﺸﺎﻫﺪﺓ‬ ‫ﺍﻹﺷﺎﺭﺓ‬ ‫ﺍﻟﺰﻣﻦ‬ ‫ﺍﳌﺸﺎﻫﺪﺓ‬ ‫ﺇﺷﺎﺭﺓ‬ ‫ﺍﻟﺰﻣﻦ‬
‫‪162‬‬ ‫‪-‬‬ ‫‪7‬‬ ‫‪155‬‬ ‫‪-‬‬ ‫‪1‬‬
‫‪172‬‬ ‫‪+‬‬ ‫‪8‬‬ ‫‪158‬‬ ‫‪-‬‬ ‫‪2‬‬
‫‪162‬‬ ‫‪-‬‬ ‫‪9‬‬ ‫‪163‬‬ ‫‪+‬‬ ‫‪3‬‬
‫‪164‬‬ ‫‪+‬‬ ‫‪10‬‬ ‫‪171‬‬ ‫‪+‬‬ ‫‪4‬‬
‫‪173‬‬ ‫‪+‬‬ ‫‪11‬‬ ‫‪153‬‬ ‫‪-‬‬ ‫‪5‬‬
‫‪181‬‬ ‫‪+‬‬ ‫‪12‬‬ ‫‪156‬‬ ‫‪-‬‬ ‫‪6‬‬
‫ﻳﻜﻮﻥ ﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﺘﺼﺎﻋﺪﻱ ﳍﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ ﳑﺜﻼﹰ ﰲ ﺍﳉﺪﻭﻝ )‪ (5‬ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(5‬ﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﺘﺼﺎﻋﺪﻱ‬
‫ﺍﳌﺸﺎﻫﺪﺓ‬ ‫ﺍﻟﺮﺗﺒﺔ‬ ‫ﺍﳌﺸﺎﻫﺪﺓ‬ ‫ﺍﻟﺮﺗﺒﺔ‬ ‫ﺍﳌﺸﺎﻫﺪﺓ‬ ‫ﺍﻟﺮﺗﺒﺔ‬
‫‪171‬‬ ‫‪9‬‬ ‫‪162‬‬ ‫‪5‬‬ ‫‪153‬‬ ‫‪1‬‬
‫‪172‬‬ ‫‪10‬‬ ‫‪162‬‬ ‫‪5‬‬ ‫‪155‬‬ ‫‪2‬‬
‫‪173‬‬ ‫‪11‬‬ ‫‪163‬‬ ‫‪7‬‬ ‫‪156‬‬ ‫‪3‬‬
‫‪181‬‬ ‫‪12‬‬ ‫‪164‬‬ ‫‪8‬‬ ‫‪158‬‬ ‫‪4‬‬

‫ﺣﻴﺚ‬
‫‪T‬‬
‫=‪m‬‬
‫‪2‬‬
‫‪m=6‬‬ ‫ﻭ‬

‫‪35‬‬
 ‫) ‪( cm + cm +1‬‬ ‫ﺑﻴﻨﻤﺎ‬
‫= ‪Md‬‬
‫‪2‬‬
‫) ‪( c5 + c 7‬‬
‫= ‪Md‬‬
‫‪2‬‬

‫‪= (162 + 163) / 2 = 162.5‬‬

‫ﰎﹼ ﺍﺳﺘﺨﺪﺍﻡ ﺍﻟﺮﺗﺒﺔ ﺍﳋﺎﻣﺴﺔ ﻭﺍﻟﺴﺎﺑﻌﺔ ﺑﺴﺒﺐ ﺗﻜﺮﺍﺭ ﺍﻷﻭﱃ ﻧﺘﻴﺠﺔ ﻭﺟﻮﺩ‬
‫ﻣﺸﺎﻫﺪﺗﲔ ﻣﺘﺴﺎﻭﻳﺘﲔ‪ .‬ﻓﺈﻋﻄﺎﺀ ﺇﺷﺎﺭﺓ ﻣﻮﺟﺒﺔ ﻟﻠﻘﻴﻢ ﺍﻷﻛﱪ ﻣﻦ ﺍﻟﻮﺳﻴﻂ ﻭﺳﺎﻟﺒﺔ‬
‫ﻟﻠﻘﻴﻢ ﺍﻷﺻﻐﺮ ﻣﻨﻪ ﺿﻤﻦ ﺍﳉﺪﻭﻝ ﺍﻷﺻﻠﻲ‪ ،‬ﳓﺼﻞ ﻋﻠﻰ ﻣﺎ ﻳﻠﻲ‪:‬‬
‫‪+++‬‬ ‫‪-‬‬ ‫‪+‬‬ ‫‪---‬‬ ‫‪++‬‬ ‫‪--‬‬

‫‪R=6‬‬ ‫ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﺗﻜﻮﻥ‬

‫ﻭﻣﻨﻪ ﻧﻘﺒﻞ‬ ‫‪m=6‬‬ ‫ﻣﻦ ﺍﳉﺪﻭﻝ ﻧﺴﺘﺨﺮﺝ ﺍﻟﻘﻴﻢ ﺍﻟﻘﺼﻮﻯ ﺍﳌﻘﺎﺑﻠﺔ‬
‫ﺍﻟﻔﺮﺿﻴﺔ ‪ H 0‬ﺇﺫﺍ ﱂ ﻳﺘﺤﻘﻖ ﺃﺣﺪ ﺍﻟﺸﺮﻃﲔ ﺍﻟﺘﺎﻟﻴﲔ‪:‬‬
‫‪R £ Rl‬‬ ‫ﺃﻭ‬ ‫‪R ³ Ru‬‬
‫ﺍﳌﻘﺎﺑﻠﺔ ﻟـ‪. m = 6‬‬ ‫‪Rl=3‬‬ ‫ﻭ‬ ‫‪Ru = 11‬‬ ‫ﺃﻳﻦ‬
‫ﺍﳌﺨﻄﻂ )‪ :(3‬ﺍﺧﺘﺒﺎﺭ ﺍﻟﺘﻮﺍﱄ‬

‫‪28‬‬ ‫‪0‬‬
‫‪RU= 11‬‬ ‫‪RL= 3‬‬

‫ﺍﻟﻘﺮﺍﺭ‪ :‬ﺣﻴﺚ ﺃﻥ ﺷﺮﻁ ﺍﻟﺮﻓﺾ ﱂ ﻳﺘﺤﻘﻖ ﻧﻘﻮﻝ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺧﺎﻟﻴﺔ ﻣﻦ‬
‫ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ .‬ﻟﻠﺘﺄﻛﺪ ﻣﻦ ﺻﺤﺔ ﻫﺬﺍ ﺍﻟﻘﺮﺍﺭ ﻧﻮﺩ ﺍﻟﺮﺟﻮﻉ ﺇﱃ ﺍﻟﺮﺳﻢ ﺍﻟﺒﻴﺎﱐ‬
‫ﺍﻟﺬﻱ ﻳﻌﻜﺴﻪ ﺍﻟﺸﻜﻞ )‪ (3‬ﺍﳌﺒﲔ ‪ -‬ﻭﺭﻏﻢ ﺍﻻﻫﺘﺰﺍﺯﺍﺕ ﺍﳌﺴﺠﻠﺔ ﰲ ﺍﳌﻨﺤﲎ‪ -‬ﺃﻥ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﲣﻀﻊ ﻟﺘﺄﺛﲑ ﻣﻨﺘﻈﻢ ﲟﻴﻞ ﻣﻮﺟﺐ ﻳﻌﻜﺲ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ‬
‫ﺍﻟﻌﺎﻡ‪.‬‬

‫‪36‬‬
‫ﻋﻠﻰ ﻫﺬﺍ ﺍﻷﺳﺎﺱ ﳝﻜﻦ ﺍﻹﺷﺎﺭﺓ ﺇﱃ ﺍﺣﺘﻤﺎﻝ ﻭﻗﻮﻉ ﺧﻄﺄ ﰲ ﺍﻟﻘﺮﺍﺭ ﻭﺑﺎﻟﺘﺎﱄ‬
‫ﻗﺪ ﻳﻜﻮﻥ ﻣﻐﻠﻄﺎﹰ ﻭﻫﻮ ﻣﺎ ﻳﺴﺘﺪﻋﻲ ﺍﳊﺬﺭ ﻋﻨﺪ ﺍﺳﺘﻌﻤﺎﻝ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ‬
‫ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﻭﻋﺪﻡ ﺍﻋﺘﺒﺎﺭﻫﺎ ﺻﺤﻴﺤﺔ ﻣﻄﻠﻘﺔ ﻭﺍﻟﺮﻛﻮﻥ ﺇﻟﻴﻬﺎ ﻭﻋﺪﻡ ﺍﻟﺒﺤﺚ ﻋﻦ‬
‫ﺍﻷﻓﻀﻞ‪.‬‬

‫ﺍﻻﻧﻌﻄﺎﻑ‪(Turning points) :‬‬ ‫ﺍ‪ -2-‬ﺍﺧﺘﺒﺎﺭ ﻧﻘﺎﻁ‬

‫ﺇﻥ ﺍﻟﺘﺴﻤﻴﺔ ﺍﻟﻮﺍﺭﺩﺓ ﺃﻋﻼﻩ ﻏﲑ ﻣﻌﱪﺓ ﺑﺸﻜﻞ ﺩﻗﻴﻖ ﻋﻠﻰ ﺍﻟﻔﻜﺮﺓ‪ ،‬ﻛﻮﻥ‬
‫ﺍﻻﺧﺘﺒﺎﺭ ﻭﰲ ﺗﻜﻮﻳﻨﻪ ﻻ ﻳﻬﺘﻢ ﺑﻨﻘﺎﻁ ﺍﻻﻧﻌﻄﺎﻑ ﲝﺪ ﺫﺍ‪‬ﺎ ﺍﻟﱵ ﺗﻌﻜﺲ ﺗﻐﲑ ﺍﲡﺎﻩ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﺇﳕﺎ ﺑﻌﺪﺩ ﻣﺮﺍﺕ ﺍﻟﺼﻌﻮﺩ ﻭﺍﻟﻨﺰﻭﻝ )‪ (up and down‬ﻟﻠﻤﻨﺤﲎ‪،‬‬
‫ﺣﻴﺚ ﺍﻹﺷﺎﺭﺓ ﺍﳌﻮﺟﺒﺔ ﺗﻌﲏ ﺍﻟﺼﻌﻮﺩ ﻭ ﺍﻟﺴﺎﻟﺒﺔ ﺗﻌﲏ ﺍﳍﺒﻮﻁ ﰲ ﺗﻠﻚ ﺍﻟﻔﺘﺮﺓ‪.‬‬
‫ﻓﺤﺴﺐ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ‪ ،‬ﺗﻌﺘﱪ ﻧﻘﻄﺔ ﺍﻧﻌﻄﺎﻑ ﺗﻠﻚ ﺍﻟﻔﺘﺮﺓ ﺍﻟﱵ ﺗﻜﻮﻥ‬
‫ﺍﻹﺷﺎﺭﺓ ﻓﻴﻬﺎ ﳐﺘﻠﻔﺔ ﻋﻦ ﺇﺷﺎﺭﺓ ﺍﻟﻔﺘﺮﺓ ﺍﻟﺴﺎﺑﻘﺔ‪ .‬ﻓﺈﺫﺍ ﻛﺎﻧﺖ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﻋﺸﻮﺍﺋﻴﺔ ﺑﺪﻭﻥ ﺍﲡﺎﻩ ﻋﺎﻡ‪ ،‬ﻓﺈﻥ ﺗﻮﺯﻳﻊ ﻋﺪﺩ ﻣﺮﺍﺕ ﺗﻐﻴ‪‬ﺮ ﺍﻹﺷﺎﺭﺓ ﻳﻜﻮﻥ ﺗﻘﺮﻳﺒﺎ‬
‫ﻃﺒﻴﻌﻴﺎﹰ ﺣﱴ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻌﻴﻨﺎﺕ ﺍﻟﺼﻐﲑﺓ‪ ،‬ﳑﺎ ﻳﻘﺘﻀﻲ ﺍﻻﺳﺘﻌﺎﻧﺔ ﲜﺪﺍﻭﻝ ﺍﻟﺘﻮﺯﻳﻊ‬
‫ﺍﻟﻄﺒﻴﻌﻲ ﻻﺳﺘﺨﺮﺍﺝ ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ‪.‬‬
‫ﳍﺬﺍ ﳝﻜﻦ ﺗﻜﻮﻳﻦ ﺍﻻﺧﺘﻴﺎﺭ ﻣﻦ ﺧﻼﻝ ﺣﺴﺎﺏ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‬
‫‪ Dy t‬ﺃﻳﻦ‪:‬‬
‫‪Dy t = y t - y t - 1‬‬ ‫)‪(2.6‬‬

‫ﺣﻴﺚ ‪ yt‬ﲤﺜﻞ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻗﻴﺪ ﺍﻻﺧﺘﺒﺎﺭ ﻣﺮﺗﺒﺔ ﺗﺮﺗﻴﺒﺎﹰ ﺯﻣﻨﻴﺎﹰ‪.‬‬

‫ﺻﻴﻐﺔ ﺍﻻﺧﺘﺒﺎﺭ‬
‫‪ :‬ﺍﻟﺴﻠﺴﻠﺔ ﻋﺸﻮﺍﺋﻴﺔ \ ﻻ ﻳﻮﺟﺪ ﺍﲡﺎﻩ ﻋﺎﻡ ‪.‬‬ ‫‪H0‬‬

‫‪ :‬ﻳﻮﺟﺪ ﺍﲡﺎﻩ ﻋﺎﻡ ‪.‬‬ ‫‪HA‬‬

‫‪37‬‬
 ‫ﺗﻜﻮﻳﻦ ﺍﻻﺧﺘﺒﺎﺭ‬
‫‪ -1‬ﺣﺴﺎﺏ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﻟﻠﺴﻠﺴﻠﺔ ﺍﳌﻌﻨﻴﺔ‪،‬‬
‫‪ -2‬ﺇﻋﻄﺎﺀ ﺇﺷﺎﺭﺓ ﻣﻮﺟﺒﺔ ﻟﻠﻔﺮﻭﻕ ﺍﳌﻮﺟﺒﺔ ﻭﺳﺎﻟﺒﺔ ﻟﻠﻔﺮﻭﻕ ﺍﻟﺴﺎﻟﺒﺔ‪.‬‬
‫ﻳﺮﻣﺰ ﳍﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﺑﺎﳊﺮﻑ‪ U‬ﻭﻫﻮ ﻋﺪﺩ ﻣﺮﺍﺕ ﺗﻐﻴ‪‬ﺮ ﺍﻹﺷﺎﺭﺓ ﰲ ‪. Dy t‬‬
‫ﻳﺴﺘﻌﻤﻞ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﰲ ﺍﳊﺎﻻﺕ ﺍﻟﱵ ﻳﻜﻮﻥ ﻓﻴﻬﺎ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺃﻛﱪ ﻣﻦ ‪.10‬‬
‫ﺇﺫﺍ ﻛﺎﻥ‪:‬‬ ‫‪H0‬‬ ‫ﺍﻟﻘـﺮﺍﺭ‪ :‬ﺭﻓﺾ‬
‫‪Z‬‬ ‫‪> Z‬‬ ‫‪a‬‬
‫‪2‬‬

‫ﺣﻴﺚ ‪ Z‬ﻣﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪U - mu‬‬ ‫)‪(2.7‬‬
‫= ‪Z‬‬
‫‪su‬‬
‫ﺃﻳﻦ‬
‫)‪2(T - 2‬‬ ‫)‪(2.8‬‬
‫= ‪mu‬‬
‫‪3‬‬

‫ﻭ‬
‫‪16 T - 29‬‬ ‫)‪(2.9‬‬
‫= ‪su‬‬
‫‪90‬‬
‫ﻣﺜﺎﻝ ‪ :6‬ﺑﺎﺳﺘﺨﺪﺍﻡ ﻣﻌﻄﻴﺎﺕ ﺍﳌﺜﺎﻝ)‪ (5‬ﻭﻣﺴﺘﻌﻴﻨﹰﺎ ‪‬ﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ‪ ،‬ﺍﳌﻄﻠﻮﺏ ﲢﺪﻳﺪ‬
‫ﻋﺸﻮﺍﺋﻴﺔ ﺳﻠﺴﻠﺔ ﺍﻻﺳﺘﻬﻼﻙ ﺍﻟﱪﻳﻄﺎﱐ ﺍﳌﻮﲰﻲ‪.‬‬
‫اﻟﺠﺪول )‪ :(6‬ﺗﺸﻜﯿﻞ إﺷﺎرات اﻻﺧﺘﺒﺎر‬
‫‪12‬‬ ‫‪11‬‬ ‫‪10‬‬ ‫‪9‬‬ ‫‪8‬‬ ‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪t‬‬
‫‪181‬‬ ‫‪173‬‬ ‫‪164‬‬ ‫‪162‬‬ ‫‪172‬‬ ‫‪162‬‬ ‫‪156‬‬ ‫‪153‬‬ ‫‪171‬‬ ‫‪163‬‬ ‫‪158‬‬ ‫‪155‬‬ ‫‪Ct‬‬
‫‪8‬‬ ‫‪9‬‬ ‫‪2‬‬ ‫‪10-‬‬ ‫‪10‬‬ ‫‪6‬‬ ‫‪3‬‬ ‫‪18-‬‬ ‫‪8‬‬ ‫‪5‬‬ ‫‪3‬‬ ‫‪....‬‬ ‫‪DCt‬‬
‫‪+‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫‪-‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫‪-‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫‪....‬‬ ‫اﻹﺷﺎرة‬

‫‪U = 5, T > 10‬‬ ‫ﺇﺫﺍ‬

‫ﺑﺘﻄﺒﻴﻖ ﺍﻟﻌﻼﻗﺘﲔ )‪ (2.8‬ﻭ )‪ (2.9‬ﺃﻋﻼﻩ ﻳﻜﻮﻥ‪:‬‬
‫‪m u = 6 . 66‬‬

‫‪s u = 1 . 34‬‬

‫‪38‬‬
 ‫‪Z = 1 . 23‬‬ ‫ﳒﺪ ﺃﻥ‪:‬‬ ‫ﺍﻟﻌﻼﻗﺔ)‪(2.7‬‬ ‫ﺑﺎﺳﺘﺨﺪﺍﻡ‬
‫ﺍﻟﻘﺮﺍﺭ‪ :‬ﻗﺒﻮﻝ ‪ H 0‬ﺃﻱ ﻋﺸﻮﺍﺋﻴﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﳌﺎﹼ ﺗﻜﻮﻥ‪:‬‬
‫‪Z < Za‬‬
‫‪2‬‬

‫‪z a = z 5 % = z 2 . 5 % = 1 . 96‬‬ ‫‪ 1. 23 < 1. 96‬ﺃﻳﻦ‬ ‫ﺃﻱ‬

‫‪2‬‬ ‫‪2‬‬

‫ﻧﻼﺣﻆ ﺃﻥ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻗﺪ ﺃﻛﺪ ﻧﺘﻴﺠﺔ ﺍﻟﻘﺮﺍﺭ ﺍﻟﺴﺎﺑﻖ ﻻﺧﺘﺒﺎﺭ ) ‪ .( R‬ﻛﻤﺎ‬

‫ﳝﻜﻦ ﺍﻹﺷﺎﺭﺓ ﺇﱃ ﺃﻥ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻗﺪ ﻳﺴﺘﻌﻤﻞ ﻟﻜﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﺍﻟﱵ ﳝﻜﻦ‬
‫ﻣﻼﺣﻈﺔ ﺃﺛﺮﻫﺎ ﻣﻦ ﺧﻼﻝ ﺍﻻﻧﺘﻈﺎﻡ ﺍﳌﺴﺠﻞ ﰲ ﺍﻹﺷﺎﺭﺍﺕ ﺍﳌﻮﺟﺒﺔ ﻭﺍﻟﺴﺎﻟﺒﺔ )ﺛﻼﺙ‬
‫ﺇﺷﺎﺭﺍﺕ ﻣﻮﺟﺒﺔ ﻣﺘﺒﻮﻋﺔ ﺑﺈﺷﺎﺭﺓ ﺳﺎﻟﺒﺔ ﰲ ﻛﻞ ﺳﻨﺔ ﻛﻤﺎ ﻳﺒﻨﻪ ﺍﻟﺴﻄﺮ ﺍﻷﺧﲑ‬
‫ﻟﻠﺠﺪﻭﻝ)‪ (6‬أﻋﻼه ‪ ،‬ﻫﺬﺍ ﰲ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﺑﻴﻨﻤﺎ ﰲ ﺍﻟﺸﻬﺮﻳﺔ ﻣﺜﻼﹰ ﳚﺐ‬
‫‪4‬‬

‫ﺍﳌﻘﺎﺭﻧﺔ ﻋﻠﻰ ﺃﺳﺎﺱ )‪ (12‬ﺇﺷﺎﺭﺓ‪ ،‬ﻓﻴﺘﻢ ﻣﻘﺎﺭﻧﺔ ﳎﻤﻮﻋﺔ ﺇﺷﺎﺭﺍﺕ ﺍﻟﺴﻨﺔ ﺍﻷﻭﱃ ﻣﻊ‬
‫ﳎﻤﻮﻋﺔ ﺇﺷﺎﺭﺍﺕ ﺍﻟﺴﻨﺔ ﺍﻟﺜﺎﻧﻴﺔ ﻭﻫﻜﺬﺍ‪.‬‬

‫)‪(Sign test‬‬ ‫ا‪ - 3-‬ﺍﺧﺘﺒﺎﺭ ﺍﻹﺷﺎﺭﺓ‬

‫ﻋﻠﻰ ﻏﺮﺍﺭ ﻣﺎ ﺳﺒﻖ‪ ،‬ﻳﻌﺘﻤﺪ ﺍﺧﺘﺒﺎﺭ ﺍﻹﺷﺎﺭﺓ ) ‪ ( V‬ﻋﻠﻰ ﺇﺷﺎﺭﺍﺕ ﺍﻟﻔﺮﻭﻕ ﻣﻦ‬
‫ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﻣﻦ ﻣﻮﺟﺒﺔ ﻭﺳﺎﻟﺒﺔ‪ ،‬ﻛﻤﺎ ﻳﻔﺘﺮﺽ ﺍﻟﺘﻮﺯﻳﻊ ﺍﻟﻌﺸﻮﺍﺋﻲ ﻟﻠﻤﻌﻄﻴﺎﺕ‪.‬‬
‫ﺻﻴﻐﺔ ﺍﻻﺧﺘﺒﺎﺭ‬
‫‪ :‬ﺍﻟﺴﻠﺴﻠﺔ ﻋﺸﻮﺍﺋﻴﺔ \ ﻻ ﻳﻮﺟﺪ ﺍﲡﺎﻩ ﻋﺎﻡ‬ ‫‪H0‬‬

‫‪ : H A‬ﻳﻮﺟﺪ ﺍﲡﺎﻩ ﻋﺎﻡ ‪.‬‬

‫ﺻﻔﺤﺔ )‪.(45‬‬ ‫‪ - 4‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ "ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﳊﺮﺓ ﻟﻜﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ"‪،‬‬

‫‪39‬‬
 ‫ﺗﻜﻮﻳﻦ ﺍﻻﺧﺘﺒﺎﺭ‬
‫ﺍ‪-‬ﲢﺪﻳﺪ ﻋﺪﺩ ﺍﻟﻔﺮﻭﻕ ﺍﳌﻮﺟﺒﺔ ‪ V‬ﻭ ﻋﺪﺩ ﺍﻟﻔﺮﻭﻕ ﻏﲑ ﺍﻟﺼﻔﺮﻳﺔ ‪. n‬‬
‫ﺏ‪ -‬ﻳﺴﺘﻌﻤﻞ ﳌﺎ ) ‪( n ³ 20 , a‬‬
‫ﺍﻟﻘـﺮﺍﺭ‪ :‬ﺭﻓﺾ ‪ H 0‬ﺇﺫﺍ ﻛﺎﻥ‪:‬‬
‫‪Z > Za‬‬
‫‪2‬‬

‫ﺣﻴﺚ ‪ Z‬ﻣﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪V - mv‬‬ ‫)‪(2.10‬‬
‫= ‪Z‬‬
‫‪sv‬‬
‫ﺑﻴﻨﻤﺎ‬
‫‪m‬‬ ‫=‬
‫‪n‬‬ ‫)‪(2.11‬‬
‫‪v‬‬
‫‪2‬‬

‫= ‪sv‬‬
‫‪n‬‬ ‫)‪(2.12‬‬
‫‪4‬‬
‫ﻣﺜﺎﻝ‪ :7‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ ﺍﳌﺘﻌﻠﻘﺔ ﲟﺘﻐﲑ ﺍﺟﺘﻤﺎﻋﻲ‪ ،‬ﺍﺧﺘﱪ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ‬
‫ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬

‫ﺍﳉﺪﻭﻝ )‪ :(7‬ﻣﻌﻄﻴﺎﺕ ﻣﺘﻐﲑ ﺍﺟﺘﻤﺎﻋﻲ‬

‫‪12‬‬ ‫‪11‬‬ ‫‪10‬‬ ‫‪9‬‬ ‫‪8‬‬ ‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪t‬‬
‫‪36‬‬ ‫‪60‬‬ ‫‪18‬‬ ‫‪7‬‬ ‫‪29‬‬ ‫‪68‬‬ ‫‪12‬‬ ‫‪12‬‬ ‫‪32‬‬ ‫‪64‬‬ ‫‪19‬‬ ‫‪10‬‬ ‫‪wt‬‬
‫‪24‬‬ ‫‪23‬‬ ‫‪22‬‬ ‫‪21‬‬ ‫‪20‬‬ ‫‪19‬‬ ‫‪18‬‬ ‫‪17‬‬ ‫‪16‬‬ ‫‪15‬‬ ‫‪14‬‬ ‫‪13‬‬ ‫‪t‬‬
‫‪30‬‬ ‫‪54‬‬ ‫‪20‬‬ ‫‪9‬‬ ‫‪20‬‬ ‫‪69‬‬ ‫‪19‬‬ ‫‪11‬‬ ‫‪50‬‬ ‫‪64‬‬ ‫‪11‬‬ ‫‪6‬‬ ‫‪wt‬‬

‫ﲝﺴﺎﺏ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﳍﺬﺍ ﺍﳌﺘﻐﲑ ﻭﻣﻦ ﰒ ﲢﺪﻳﺪ ﻋﺪﺩ‬

‫ﺍﻹﺷﺎﺭﺍﺕ ﺍﳌﻮﺟﺒﺔ ﻭﻛﺬﺍ ﻋﺪﺩ ﺍﻟﻔﺮﻭﻕ ﻏﲑ ﺍﻟﺼﻔﺮﻳﺔ‪ ،‬ﻧﺴﺘﻨﺘﺞ ﻣﺎ ﻳﻠﻲ‪:‬‬

‫‪40‬‬
 ‫ﺍﳉﺪﻭﻝ )‪ :(8‬ﺣﺴﺎﺏ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‬
‫‪12‬‬ ‫‪11‬‬ ‫‪10‬‬ ‫‪9‬‬ ‫‪8‬‬ ‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪t‬‬
‫‪-24‬‬ ‫‪42‬‬ ‫‪11‬‬ ‫‪-22‬‬ ‫‪-39‬‬ ‫‪56‬‬ ‫‪0‬‬ ‫‪-20‬‬ ‫‪-32‬‬ ‫‪45‬‬ ‫‪9‬‬ ‫‪-‬‬ ‫‪Dw t‬‬
‫‪24‬‬ ‫‪23‬‬ ‫‪22‬‬ ‫‪21‬‬ ‫‪20‬‬ ‫‪19‬‬ ‫‪18‬‬ ‫‪17‬‬ ‫‪16‬‬ ‫‪15‬‬ ‫‪14‬‬ ‫‪13‬‬ ‫‪t‬‬
‫‪-24‬‬ ‫‪34‬‬ ‫‪11‬‬ ‫‪-11‬‬ ‫‪-49‬‬ ‫‪50‬‬ ‫‪8‬‬ ‫‪-39‬‬ ‫‪-14‬‬ ‫‪53‬‬ ‫‪5‬‬ ‫‪-30‬‬ ‫‪Dw t‬‬

‫‪V = 11‬‬ ‫ﺍ‪ -‬ﻋﺪﺩ ﺍﻹﺷﺎﺭﺍﺕ ﺍﳌﻮﺟﺒﺔ‬

‫‪n = 22‬‬ ‫ﺏ‪ -‬ﻋﺪﺩ ﺍﻟﻔﺮﻭﻕ ﻏﲑ ﺍﻟﺼﻔﺮﻳﺔ‬
‫ﺇﺫﺍ‬
‫‪V-m‬‬ ‫‪11 - m‬‬
‫=‪z‬‬ ‫= ‪v‬‬ ‫‪v‬‬
‫‪s‬‬ ‫‪s‬‬
‫‪v‬‬ ‫‪v‬‬
‫ﺑﻴﻨﻤﺎ‬
‫‪n‬‬
‫= ‪mv‬‬ ‫‪= 11‬‬
‫‪2‬‬
‫ﻭ‬
‫‪22‬‬
‫= ‪sv‬‬ ‫‪= 2 . 345‬‬
‫‪4‬‬
‫ﺃﻳﻦ ﻭﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﺍﳋﺎﺻﺔ‬
‫‪11 - 11‬‬
‫=‪z‬‬ ‫‪= 0 . 00‬‬
‫‪2 . 345‬‬
‫ﺍﻟﻘـﺮﺍﺭ‪ :‬ﺇﺫﺍ ﻛﺎﻧﺖ ‪ a = 5% :‬ﻓﺈﻥ‬
‫‪5‬‬

‫‪Z‬‬ ‫‪a‬‬ ‫‪= 1 . 96‬‬

‫‪2‬‬

‫ﻛﻮﻥ ﺍﻹﺣﺼﺎﺀﺓ ﺍﶈﺴﻮﺑﺔ ﺃﻗﻞ ﻣﻦ ﺗﻠﻚ ﺍ‪‬ﺪﻭﻟﺔ ﺃﻋﻼﻩ‪ ،‬ﻳﺘﻢ ﻗﺒﻮﻝ ﺍﻟﻔﺮﺿﻴﺔ‬
‫‪ H 0‬ﺍﻟﻨﺎﺻﺔ ﻋﻠﻰ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﻋﺸﻮﺍﺋﻴﺔ ﲝﺘﺔ‪ ،‬ﺃﻱ ﺧﺎﻟﻴﺔ ﻣﻦ ﻣﺮﻛﺒﺔ ﺍﲡﺎﻩ ﻋﺎﻡ‪.‬‬

‫‪ -5‬ﺃﻧﻈﺮ ﺟﺪﻭﻝ ﻣﺴﺎﺣﺎﺕ ﺍﻟﺘﻮﺯﻳﻊ ﺍﻟﻄﺒﻴﻌﻲ ﺍﳌﻌﻴﺎﺭﻱ‪.‬‬

‫‪41‬‬
‫ﻧﻈﺮﺍﹰ ﳌﺎ ﺫﻛﺮ ﺳﺎﺑﻘﺎﹰ ﺣﻮﻝ ﺍﻟﻘﻮﺓ ﺍﻹﺣﺼﺎﺋﻴﺔ ﳍﺬﻩ ﺍﳌﻘﺎﻳﻴﺲ‪ ،‬ﻧﺘﺮﻙ ﺍﻟﺘﻌﻠﻴﻖ‬
‫ﻟﻨﺘﺎﺋﺞ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﻟﺘﺎﻟﻴﺔ ﻭﺍﻟﱵ ﺗﻌﺘﱪ ﺃﻛﺜﺮ ﺃﳘﻴﺔ ﻣﻦ ﺣﻴﺚ ﺍﻟﻘﻮﺓ ﺍﻹﺣﺼﺎﺋﻴﺔ‬
‫ﻟﻜﺸﻒ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬
‫‪Daniels' test‬‬ ‫ا‪ - 4-‬ﺍﺧﺘﺒﺎﺭ ﺩﺍﻧﻴﺎﻝ‬
‫ﻳﻌﺘﱪ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﺃﻗﻮﻯ ﻣﻌﻴﺎﺭ ﰲ ﳎﻤﻮﻋﺘﻪ ﺍﳌﺬﻛﻮﺭﺓ ﺳﺎﺑﻘﺎﹰ‪ ،‬ﻭﻫﻮ ﻳﺴﺘﻌﲔ‬
‫ﲟﻌﺎﻣﻞ ﺍﻻﺭﺗﺒﺎﻁ ﻟﺴﺒﲑﻣﺎﻥ " ‪ ."Spearman Correlation Coefficient‬ﻳﻌﺘﻤﺪ‬
‫ﻫﺬﺍ ﺍﳌﻌﺎﻣﻞ ﻋﻠﻰ ﻗﻴﺎﺱ ﺍﻻﺭﺗﺒﺎﻁ ﺍﳋﻄﻲ ﺑﲔ ﺗﺮﺗﻴﺒﲔ‪ ،‬ﺍﻟﺮﺗﱯ )ﺍﻟﺘﺼﺎﻋﺪﻱ( ‪Rt‬‬
‫ﻭﺍﻟﺰﻣﲏ ) ‪ ( t‬ﺃﻱ ﻭﺑﺘﻌﺒﲑ ﺭﻳﺎﺿﻲ‪:‬‬
‫) ‪Rt = f (t‬‬

‫‪t = 1, T‬‬ ‫ﻭ‬ ‫‪Rt = 1, T‬‬ ‫ﺣﻴﺚ‬

‫ﻭﻣﻨﻪ ﻓﻤﻌﺎﻣﻞ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﻨﻈﺮﻱ ﻫﺬﺍ ﻳﻌﺮ‪ ‬ﻑ ﺑـ‪:‬‬

‫) ‪cov( R t , t‬‬ ‫)‪(2.13‬‬

‫= ‪rs‬‬
‫) ‪var( R t ). var( t‬‬
‫ﺃﻳﻦ ) ‪ var( R t ) = var( t‬ﻭﻳﻌﺎﺩﻻﻥ ) ‪ ، ( T - 1‬ﺣﻴﺚ ﻳﻜﻮﻥ ﰲ ﺣﺎﻟﺔ‬
‫‪2‬‬

‫‪12‬‬
‫ﺍﻟﻌﻴﻨﺔ )‪ (sample‬ﻭ ﺑﺴﻠﺴﻠﺔ ﻏﲑ ﻣﻜﺮﺭﺓ ﺍﳌﺸﺎﻫﺪﺍﺕ‪ ،‬ﻣﻌﻄﻰ ﺑـ‪:‬‬
‫‪T‬‬ ‫_‬ ‫_‬ ‫_‬
‫‪å( R‬‬
‫‪t =1‬‬
‫‪t‬‬ ‫) ‪- R )( t - t‬‬
‫)‪(2.14‬‬
‫= ‪rs‬‬ ‫‪T‬‬ ‫_‬
‫) ‪å( t - t‬‬
‫‪t =1‬‬
‫‪2‬‬

‫ﺣﻴﺚ ﻭﻣﺮﺍﻋﺎﺓ ﻟﻌﺪﻡ ﺗﻜﺮﺍﺭ ﺍﻟﺮﺗﺐ‪ ،‬ﻧﻌﻮﺽ ﺗﻠﻚ ﺍﳌﻜﺮﺭﺓ ﺍﳌﺘﺴﺎﻭﻳﺔ ﺑﻮﺳﻄﻬﺎ‬
‫ﺍﳊﺴﺎﰊ‪ .‬ﻭﻟﺘﻮﺿﻴﺢ ﻫﺬﻩ ﺍﻟﻔﻜﺮﺓ ﻧﺴﺘﻌﲔ ﺑﺎﳉﺪﻭﻝ )‪ (5‬ﺻﻔﺤﺔ )‪ (23‬ﺣﻴﺚ ﻋﻮﺽ‬
‫ﺗﻜﺮﺍﺭ ﺍﻟﺮﺗﺒﺔ ﺍﳋﺎﻣﺴﺔ ﳕﻸ ﺍﳋﺎﻧﺘﲔ ﺍﳌﻨﺎﺳﺒﺘﲔ ﺑﺎﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﳍﻤﺎ‪ ،‬ﲟﻌﲎ ﺍﻟﺮﺗﺒﺔ‬

‫‪42‬‬
‫ﺍﳋﺎﻣﺴﺔ ﺯﺍﺋﺪﺍﹰ ﺍﻟﺴﺎﺩﺳﺔ ﺗﻘﺴﻴﻢ ﺍﺛﻨﲔ ﺃﻱ )‪ .(5.5‬ﻟﺘﺒﺴﻴﻂ ﻫﺬﺍ ﺍﳌﻌﺎﻣﻞ ﳝﻜﻦ‬
‫ﺇﻋﻄﺎﺀﻩ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﻨﻬﺎﺋﻲ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪T‬‬
‫‪6å d t2‬‬
‫‪rs = 1 -‬‬ ‫‪t =1‬‬ ‫)‪(2.15‬‬
‫)‪T (T - 1‬‬
‫‪2‬‬

‫‪T‬‬
‫ﺣﻴﺚ ‪ å d t2‬ﳝﺜﻞ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﻔﺮﻕ ﺑﲔ ﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﺘﺼﺎﻋﺪﻱ ﻭﺍﻟﺰﻣﲏ ﺃﻱ‬
‫‪t =1‬‬

‫‪.‬‬ ‫‪-‬‬ ‫‪1‬‬ ‫‪£‬‬ ‫‪r‬‬ ‫‪£‬‬ ‫‪1‬‬ ‫ﻓﺈﻥ‬ ‫ﺧﻄﻲ‬

‫‪s‬‬ ‫ﺍﺭﺗﺒﺎﻁ‬ ‫ﻣﻌﺎﻣﻞ‬ ‫‪r‬‬‫‪s‬‬ ‫ﻭﻛﻮﻥ‬ ‫) ‪d t = ( Rt - t‬‬

‫ﺻﻴﻐﺘﻪ‪:‬‬
‫‪ :‬ﺍﻟﺴﻠﺴﻠﺔ ﻋﺸﻮﺍﺋﻴﺔ \ ﻻ ﻳﻮﺟﺪ ﺍﲡﺎﻩ ﻋﺎﻡ‪.‬‬ ‫‪H‬‬ ‫‪0‬‬

‫‪ :‬ﻳﻮﺟﺪ ﺍﲡﺎﻩ ﻋﺎﻡ‪.‬‬ ‫‪HA‬‬

‫ﻭﺣﺴﺐ ﺣﺠﻢ ﺍﻟﻌﻴﻨﺔ‬ ‫ﺭﻓﺾ ‪H 0‬‬ ‫ﺍﻟﻘﺮﺍﺭ‪ :‬ﻓﺒﻌﺪ ﺣﺴﺎﺏ ﻣﻌﺎﻣﻞ ﺍﻻﺭﺗﺒﺎﻁ ‪ ، rs‬ﻳﺘﻢ‬
‫ﳌﺎ ﻳﻜﻮﻥ‪:‬‬
‫) ‪( T 30‬‬ ‫ﺃ‪ -‬ﰲ ﺍﻟﻌﻴﻨﺎﺕ ﺍﻟﺼﻐﲑﺓ‬
‫‪rs > ra‬‬
‫‪2‬‬

‫)‪(T > 30‬‬ ‫ﺏ‪ -‬ﰲ ﺍﻟﻌﻴﻨﺎﺕ ﺍﻟﻜﺒﲑﺓ‬

‫‪Z > Za‬‬
‫‪2‬‬
‫ﺣﻴﺚ‬
‫‪rs - m rs‬‬
‫= ‪Z‬‬ ‫)‪(2.16‬‬
‫‪s rs‬‬
‫ﺃﻳﻦ‬
‫‪6‬‬

‫‪m rs = 0‬‬ ‫)‪(2.17‬‬

‫‪6-‬‬ ‫‪Problèmes résolus de statistique mathématique, N.Caron, P.Tassi p.173.‬‬

‫‪43‬‬
 ‫ﻭ‬
‫‪1‬‬ ‫)‪(2.18‬‬
‫= ‪s rs‬‬
‫‪T -1‬‬

‫= ‪Z‬‬
‫‪rs‬‬
‫‪= rs T - 1‬‬ ‫)‪(2.19‬‬ ‫ﺑﺎﻟﺘﻌﻮﻳﺾ‬
‫‪s rs‬‬

‫ﻣﺜﺎﻝ ‪ :8‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﻧﻔﺲ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﻟﻮﺍﺭﺩﺓ ﰲ ﺍﳌﺜﺎﻝ)‪ (5‬ﺍﳉﺪﻭﻝ )‪ (4‬ﺻﻔﺤﺔ‬

‫)‪ (32‬ﺍﳌﺘﻌﻠﻘﺔ ﺑﺎﻻﺳﺘﻬﻼﻙ ﺍﳌﻮﲰﻲ ﺍﻟﱪﻳﻄﺎﱐ‪ ،‬ﺃﺟﺐ ﻋﻠﻰ ﻧﻔﺲ ﺍﻟﺴﺆﺍﻝ‬
‫ﻭﺑﺎﺳﺘﻌﻤﺎﻝ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﺍﳌﺘﻤﺜﻞ ﰲ ﲢﺪﻳﺪ ﻣﺪﻯ ﻋﺸﻮﺍﺋﻴﺔ ﻫﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬

‫ﻟﻺﺟﺎﺑﺔ ﻋﻠﻰ ﺍﻟﺴﺆﺍﻝ ﻧﻘﻮﻡ ﲝﺴﺎﺏ ﻣﻌﺎﻣﻞ ﺍﻻﺭﺗﺒﺎﻁ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪T‬‬
‫‪6å d t2‬‬
‫‪rs = 1 -‬‬ ‫‪t =1‬‬

‫)‪T (T 2 - 1‬‬

‫اﻟﺠﺪول )‪ :(9‬ﺣﺴﺎب ﻓﺮوق وﻣﺮﺑﻌﺎت ﻓﺮوق اﻻﺧﺘﺒﺎر‬

‫‪12‬‬ ‫‪11‬‬ ‫‪10‬‬ ‫‪9‬‬ ‫‪8‬‬ ‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫اﻟﺰﻣﻦ‬
‫‪12‬‬ ‫‪11‬‬ ‫‪8‬‬ ‫‪5.5‬‬ ‫‪10‬‬ ‫‪5.5‬‬ ‫‪3‬‬ ‫‪1‬‬ ‫‪9‬‬ ‫‪7‬‬ ‫‪4‬‬ ‫‪2‬‬ ‫اﻟﺮﺗﺒﺔ‬
‫‪0‬‬ ‫‪0‬‬ ‫‪2‬‬ ‫‪3.5‬‬ ‫‪2-‬‬ ‫‪1.5‬‬ ‫‪3‬‬ ‫‪4‬‬ ‫‪5-‬‬ ‫‪4-‬‬ ‫‪2-‬‬ ‫‪1-‬‬ ‫‪d‬‬

‫‪0‬‬ ‫‪0‬‬ ‫‪4‬‬ ‫‪12.25‬‬ ‫‪4‬‬ ‫‪2.25‬‬ ‫‪9‬‬ ‫‪16‬‬ ‫‪25‬‬ ‫‪16‬‬ ‫‪4‬‬ ‫‪1‬‬ ‫‪d2‬‬
‫‪åd‬‬
‫‪2‬‬
‫‪t‬‬ ‫‪= 1 + 4 + . . . . + 0 = 93 . 5‬‬ ‫ﺃﻳﻦ‬
‫ﻭﻣﻨﻪ ﻭﺑﺎﻟﺘﻌﻮﻳﺾ‬
‫) ‪6 ( 93 . 5‬‬
‫‪rs = 1 -‬‬ ‫‪= 0. 67‬‬
‫) ‪12 ( 144 - 1‬‬

‫ﺍﻟﻘﺮﺍﺭ‪:‬‬
‫ﻳﺘﻢ ﺭﻓﺾ ﻓﺮﺿﻴﺔ‬ ‫‪ra = r2.5% = 0.5804%‬‬ ‫ﻓﺈﻥ‬ ‫‪T = 12 , a = 5 %‬‬ ‫ﻓﺈﺫﺍ ﻛﺎﻧﺖ‬
‫‪2‬‬
‫ﺍﻟﻌﺪﻡ ﻭﻣﻨﻪ ﻗﺒﻮﻝ ﺍﻟﻔﺮﺿﻴﺔ ﺍﻟﺒﺪﻳﻠﺔ ﺍﻟﱵ ﺗﻨﺺ ﻋﻠﻰ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﻫﺬﺍ‬

‫‪44‬‬
 ‫ﻳﻌﲏ ﺿﻤﻨﻴﺎ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺫﺍﺕ ﻣﺮﻛﺒﺘﲔ‪ ،‬ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪.‬‬
‫ﻣﺜﺎﻝ ‪ : 9‬ﺑﺎﻓﺘﺮﺍﺽ ﺗﻮﻓﺮ ﺳﻠﺴﻠﺔ ﺯﻣﻨﻴﺔ ﺫﺍﺕ )‪ (75‬ﻣﺸﺎﻫﺪﺓ‪ ،‬ﻭ ﺑﻌﺪ ﺍﳊﺴﺎﺏ ﻭﻓﻖ‬
‫ﺍﻟﻄﺮﻳﻘﺔ ﺍﻟﺴﺎﺑﻘﺔ ﻗﺪﺭﺕ ﻗﻴﻤﺔ ﻣﻌﺎﻣﻞ ﺍﻻﺭﺗﺒﺎﻁ ﺑـ ‪ . rs = 0.51‬ﺍﳌﻄﻠﻮﺏ ﻛﺸﻒ‬
‫ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﺍﻟﻜﺸﻒ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﺑﺎﺳﺘﺨﺪﺍﻡ ﻃﺮﻳﻘﺔ ﺍﻟﻌﻴﻨﺎﺕ ﺍﻟﻜﺒﲑﺓ ﺃﻱ‪:‬‬
‫‪Z > Za‬‬
‫‪2‬‬

‫ﺣﻴﺚ‬
‫‪rs‬‬
‫= ‪Z‬‬ ‫‪= rs‬‬ ‫‪T -1‬‬
‫‪s rs‬‬
‫ﻭﻣﻨﻪ‬
‫‪Z = 0 . 51 74 = 4 . 38‬‬

‫ﻛﻤﺎ ﻭﺭﺩ ﺿﻤﻦ ﺍﺧﺘﺒﺎﺭ‬ ‫‪a = 5%‬‬ ‫ﳌﺎ‬ ‫‪Z a = 1 . 96‬‬ ‫ﺗﻜﻮﻥ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ‬
‫‪2‬‬
‫ﻟﻠﺘﻮﺯﻳﻊ ﺍﻟﻄﺒﻴﻌﻲ ﺍﳌﻌﻴﺎﺭﻱ ﺍﻟﻮﺍﺭﺩ ﺿﻤﻦ‬ ‫ﺟﺪﻭﻝ)‪(3‬‬ ‫ﺍﻟﺘﻮﺍﱄ ﺃﻭ ﺑﺎﻟﺮﺟﻮﻉ ﺇﱃ‬
‫ﺍﳌﻼﺣﻖ‪.‬‬
‫ﺑﻌﺪ ﻣﻘﺎﺭﻧﺔ ﺍﻟﻘﻴﻤﺘﲔ ﺍﶈﺴﻮﺑﺔ ﻭﺍ‪‬ﺪﻭﻟﺔ‪ ،‬ﻳﺘﻢ ﺭﻓﺾ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ﻭﻣﻨﻪ ﻗﺒﻮﻝ‬
‫ﻓﺮﺿﻴﺔ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬
‫)‪(Parametric Tests‬‬ ‫ﺏ ‪ -‬ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﻏﲑ ﺍﳊﺮﺓ‪:‬‬
‫ﺗﻌﺘﻤﺪ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻋﻠﻰ ﺍﻓﺘﺮﺍﺽ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﲡﺎﻩ ﻋﺎﻡ ﰲ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﺍﻟﺰﻣﻨﻴﺔ ﺇﺿﺎﻓﺔﹰ ﺇﱃ ﺍﻟﻌﺸﻮﺍﺋﻴـﺔ ﺑﺘـﻮﺯﻳـﻊ ﺍﺣﺘﻤﺎﱄ ﻟﻸﺧﻄﺎﺀ ﻣﻌﺮﻭﻑ ﺃﻱ‪:‬‬

‫) ‪y t = f ( t, u t‬‬

‫) ‪u t ® N ( 0, s 2‬‬ ‫ﲝﻴﺚ‬

‫‪45‬‬
‫ﲟﻌﲎ ﺃﻥ ﺍﳋﻄﺄ ﺍﻟﻌﺸﻮﺍﺋﻲ ‪ ut‬ﻳﺘﺒﻊ ﺗﻮﺯﻳﻌﺎﹰ ﻃﺒﻴﻌﻴﺎﹰ ﺑﻮﺳﻂ ﻣﻌﺪﻭﻡ)‪ (0‬ﻭﺗﺒﺎﻳﻦ‬
‫ﺛﺎﺑﺖ ‪ . s 2‬ﻭﺑﻌﺪ ﲢﺪﻳﺪ ﺷﻜﻞ ﺍﻟﺪﺍﻟﺔ )‪ f (.‬ﺃﻋﻼﻩ‪ ،‬ﻳﺘﻢ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺘﻴﻬﺎ ﻛﻤﺎ‬
‫ﺳﻨﺮﻯ ﰲ ﺍﻟﻔﺼﻞ ﺍﳌﻮﺍﱄ‪ ،‬ﰒ ﺍﺧﺘﺒﺎﺭ ﻣﻌﻨﻮﻳﺔ ﻣﻌﻠﻤﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻓﻘﻂ ﺑﺎﺳﺘﻌﻤﺎﻝ‬
‫ﻣﻘﻴﺎﺱ ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ ﺃﻭ ﺇﺣﺼﺎﺀﺓ ﺳﺘﻴﻮﺩﻧﺖ )‪...(Student- t‬ﺍﱁ‪.‬‬

‫‪ -2-2-2‬ﻛﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ‬
‫ﰲ ﻛﺜﲑ ﻣﻦ ﺍﳊﺎﻻﺕ ﳝﻜﻦ ﻛﺸﻒ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻜﻞ ﺑﺴﺎﻃﺔ ﻋﻨﺪ ﻣﻌﺮﻓﺔ‬
‫ﻣﻮﺿﻮﻉ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻓﻴﻤﻜﻦ ﻣﺴﺒﻘﺎ ﺗﻮﻗﻊ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﻓﺼﻠﻴﺔ )ﺍﻟﺪﻭﺭﻳﺔ( ﰲ‬
‫ﻣﻌﻄﻴﺎﺕ ﻓﺼﻠﻴﺔ ﺧﺎﺻﺔ ﻣﺜﻼﹰ ﲟﻮﺿﻮﻉ ﺍﻟﻄﻠﺐ ﻋﻠﻰ ﻏﺎﺯﺍﺕ ﺍﻟﺘﺪﻓﺌﺔ ) ‪GAS,GAS‬‬
‫‪ (OIL‬ﺣﻴﺚ ﻳﺰﺩﺍﺩ ﺍﻟﻄﻠﺐ ﻋﻠﻴﻬﺎ ﲝﺪﺓ ﰲ ﺍﻟﻔﺼﻮﻝ ﻭﺍﳌﻨﺎﻃﻖ ﺍﻟﺒﺎﺭﺩﺓ ﻛﺎﳍﻀﺎﺏ‬
‫ﺍﻟﻌﻠﻴﺎ ﰲ ﺍﳉﺰﺍﺋﺮ‪ ،‬ﺑﻴﻨﻤﺎ ﻭﰲ ﺍﳉﺰﺍﺋﺮ ﺩﺍﺋﻤﺎ ﻭﻋﻜﺲ ﺍﻟﻄﻠﺐ ﻋﻠﻰ ﺍﻟﻐﺎﺯ‪ ،‬ﻳﻨﻘﺺ‬
‫ﺍﻟﻄﻠﺐ ﻋﻠﻰ ﺍﻟﻜﻬﺮﺑﺎﺀ ﰲ ﺍﻟﻔﺼﻮﻝ ﺍﻟﺒﺎﺭﺩﺓ ﻭﻳﺰﺩﺍﺩ ﰲ ﺗﻠﻚ ﺍﳊﺎﺭﺓ ﻧﻈﺮﺍﹰ ﻟﻼﺳﺘﻌﻤﺎﻝ‬
‫ﺍﳌﻜﺜﻒ ﻷﺟﻬﺰﺓ ﺍﻟﺘﱪﻳﺪ‪ .‬ﻛﻤﺎ ﻳﺴﺘﺨﺪﻡ ﺍﻟﺒﺎﺣﺚ ﳎﻤﻮﻋﺔ ﻣﻌﺘﱪﺓ ﻣﻦ ﺍﻷﺩﻭﺍﺕ‬
‫ﺍﻹﺣﺼﺎﺋﻴﺔ ﺍﻻﺧﺘﻴﺎﺭﻳﺔ ﳌﻌﺮﻓﺔ ﻣﺪﻯ ﺗﻮﺍﺟﺪ ﺍﳌﺮﻛﺒﺔ ﰲ ﺍﻟﺴﻠﺴﻠﺔ‪ .‬ﻋﻠﻰ ﻏﺮﺍﺭ‬
‫ﺍﺧﺘﺒﺎﺭﺍﺕ ﻛﺴﻒ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﳝﻜﻦ ﺍﻻﺳﺘﻌﺎﻧﺔ ﲟﺠﻤﻮﻋﱵ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﳊﺮﺓ ﻭ‬
‫ﻏﲑ ﺍﳊﺮﺓ ﻟﻜﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ ﰲ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺍﳌﺪﺭﻭﺳﺔ‪.‬‬

‫ﺍﻟﺸﻜﻞ )‪ :(8‬ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ‬

‫‪46‬‬
‫ﻛﻤﺎ ﳝﻜﻦ ﻛﺸﻒ ﻫﺬﻩ ﺍﳌﺮﻛﺒﺔ ﻣﻦ ﺧﻼﻝ ﺍﻟﺮﺳﻢ ﺍﻟﺒﻴﺎﱐ‪ ،‬ﺣﻴﺚ ﻧﺴﺠﻞ‬
‫ﻗﻤﻢ ﻭﺍﳔﻔﺎﺿﺎﺕ ﰲ ﻓﺘﺮﺍﺕ ﻣﻨﺘﻈﻤﺔ ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﺳﺎﺑﻘﺎﹰ ﻭﻛﻤﺎ ﻫﻮ ﻣﺒﲔ ﰲ ﺍﻟﺸﻜﻞ‬
‫)‪ (8‬ﺃﻋﻼﻩ‪ ،‬ﺍﻟﺬﻱ ﻳﻌﻜﺲ ﻗﻤﺔ ﰲ ﺍﻟﻔﺼﻞ ﺍﻟﺜﺎﻟﺚ ﻟﻜﻞ ﺳﻨﺔ‪ .‬ﻭﺭﻏﻢ ﺫﻟﻚ‪ ،‬ﻓﺈﻧﻪ ﻗﺪ‬
‫ﻳﺘﻌﺬﺭ ﻛﺸﻔﻬﺎ ﰲ ﺑﻌﺾ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺸﺪﻳﺪﺓ ﺍﻟﺘﺬﺑﺬﺏ‪ ،‬ﻭﺧﺎﺻﺔ ﻋﻨﺪ ﺗﻮﻓﺮ ﳎﻤﻮﻋﺔ‬
‫ﻣﻬﻤﺔ ﻣﻦ ﺍﳌﻌﻄﻴﺎﺕ‪ ،‬ﺃﻭ ﻗﺪ ﳓﺘﺎﺝ ﺇﱃ ﺩﻟﻴﻞ ﺃﻛﺜﺮ ﻗﻮﺓ ﻭﺑﺮﻫﺎﻥ‪ ،‬ﻭﻣﻨﻪ ﻭﰲ ﻫﺬﻩ‬
‫ﺍﻟﻈﺮﻭﻑ ﻧﻠﺠﺄ ﺇﱃ ﺍﺳﺘﻌﻤﺎﻝ ﺑﻌﺾ ﺍﳌﻘﺎﻳﻴﺲ ﺍﻹﺣﺼﺎﺋﻴﺔ ﻟﻜﺸﻔﻬﺎ‪.‬‬

‫‪Non-parametric tests‬‬ ‫ﺍ‪ -‬ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﳊﺮﺓ ﻟﻜﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ‬

‫ﻣﻦ ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ‪ ،‬ﻟﻘﺪ ﰎﹶ ﺍﺳﺘﻌﻤﺎﻝ ﺍﺧﺘﺒﺎﺭﻱ ‪ U‬ﻭ ‪ V‬ﻟﻜﺸﻒ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ‬
‫ﺍﻟﻌﺎﻡ‪ ،‬ﺇﻻ ﺃﻧﻨﺎ ﻧﻘﺘﺼﺮ ﰲ ﻫﺬﺍ ﺍﳌﻘﺎﻡ‪ ،‬ﻭﻟﻜﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﺑﺪﺭﺍﺳﺔ ﺇﺷﺎﺭﺍﺕ ﺍﻟﻔﺮﻭﻕ‬
‫ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﻓﻘﻂ ﳌﻌﻄﻴﺎﺕ ﻣﻮﲰﻴﺔ ﻣﺜﻼﹰ‪ ،‬ﻭﺍﻟﺒﺤﺚ ﰲ ﻣﺪﻯ ﺍﻧﺘﻈﺎﻣﻬﺎ ﻛﻠﻴﺎﹰ‬
‫)‪ ( + - + + ، + - + + ،+ - + +‬ﺃﻭ ﺟﺰﺋﻴﺎ )‪+ ، - - + + ، - + + +‬‬
‫‪ ،( - + +‬ﺣﻴﺚ ﻧﺴﺠﻞ ﰲ ﻫﺬﺍ ﺍﻷﺧﲑ ﺇﺷﺎﺭﰐ )‪ (+‬ﰲ ﺑﺪﺍﻳﺔ ﻛﻞ ﺳﻨﺔ ﻣﺜﻼﹰ‬
‫ﻭﺇﺷﺎﺭﺓ )‪ (-‬ﰲ ﺍﻟﻔﺼﻞ ﺍﻟﺮﺍﺑﻊ ﻣﻦ ﻛﻞ ﺳﻨﺔ‪ .‬ﺑﻴﻨﻤﺎ ﺳﻨﺘﻌﺮﺽ ﺍﻵﻥ ﺇﱃ ﺍﺧﺘﺒﺎﺭ‬
‫‪ Kruskall-Wallis‬ﻭﻛﻴﻔﻴﺔ ﺗﻄﺒﻴﻘﻪ‪ ،‬ﻭﺍﳌﺴﺘﻌﻤﻞ ﺧﺼﻴﺼﺎ ﻟﻜﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ ﻓﻘﻂ‪.‬‬

‫ﺻﻴﻐﺔ ﺍﻻﺧﺘﺒﺎﺭ‬

‫‪47‬‬
 ‫‪ : H‬ﻻ ﺗﻮﺟﺪ ﻓﺼﻠﻴﺔ‪ /‬ﺳﻠﺴﻠﺔ ﻋﺸﻮﺍﺋﻴﺔ‬ ‫‪0‬‬

‫‪ : H A‬ﺗﻮﺟﺪ ﻓﺼﻠﻴﺔ‬
‫ﻭﻋﻼﻗﺘﻪ ﻣﻌﻄﺎﺓ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺮﻳﺎﺿﻲ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪12 é p Ri2 ù‬‬ ‫)‪(2.20‬‬
‫= ‪KW‬‬ ‫)‪êå ú - 3(T + 1) ® c ( p-1‬‬
‫‪2‬‬

‫‪T (T + 1) ë i=1 ni û‬‬

‫ﺣﻴﺚ‪:‬‬
‫‪ R i‬ﲤﺜﻞ ﳎﻤﻮﻉ ﺭﺗﺐ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﳌﻘﺎﺑﻠﺔ ﻟﻠﻔﺼﻞ ‪i‬‬
‫‪ n i‬ﲤﺜﻞ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﳌﻘﺎﺑﻠﺔ ﻟﻠﻔﺼﻞ ‪i‬‬
‫‪ p‬ﺗﻌﱪ ﻋﻦ ﺍﻟﺪﻭﺭﺓ )‪ ،(period‬ﻭﻫﻲ ﺗﻌﺎﺩﻝ )‪ (4‬ﰲ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ‬
‫ﻭ)‪ (12‬ﰲ ﺍﻟﺸﻬﺮﻳﺔ ﻭﻫﻜﺬﺍ‪.‬‬
‫ﻓﺈﺫﺍ ﻛﺎﻥ ﻛﻞ ‪ ni‬ﺃﻛﱪ ﻣﻦ )‪ (5‬ﻭﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ﺻﺤﻴﺤﺔ‪ ،‬ﳝﻜﻦ ﺇﻇﻬﺎﺭ ﺃﻥ‬
‫‪ KW‬ﻳﺘﺒﻊ ﺗﻘﺮﻳﺎﹰ ﺗﻮﺯﻳﻊ ‪ c 2‬ﺑـ ) ‪ ( p - 1‬ﺩﺭﺟﺔ ﺣﺮﻳﺔ‪.‬‬
‫ﺇﺫﺍ ﻛﺎﻥ‪:‬‬ ‫‪H‬‬ ‫‪0‬‬ ‫ﺍﻟﻘﺮﺍﺭ‪ :‬ﺭﻓﺾ‬
‫)‪KW > c 2( p -1‬‬
‫ﺇﺿﺎﻓﺔ ﻟﺸﺮﻁ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ﻭﺣﱴ ﻻ ﻳﻜﻮﻥ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻣﻐﻠﻄﺎﹰ‪ ،‬ﳚﺐ‬
‫ﺇﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﻗﺒﻞ ﳏﺎﻭﻟﺔ ﺍﻟﻜﺸﻒ ﻋﻦ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ‪.‬‬
‫ﻧﺸﲑ ﺇﱃ ﺃﻧﻪ ﳝﻜﻦ ﺇﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺑﻄﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ‬
‫ﺍﻷﻭﱃ ﺃﻭ ﻋﻦ ﻃﺮﻳﻖ ﺗﻘﻨﻴﺔ ﺍﻻﳓﺪﺍﺭ ﻛﻤﺎ ﺳﻨﺮﻯ ﻓﻴﻤﺎ ﺑﻌﺪ‪.‬‬
‫ﻣﺜﺎﻝ ‪ :10‬ﻟﺪﻳﻨﺎ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﻔﺼﻠﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ ﺣﻮﻝ ﻣﺘﻐﲑ ﺍﺟﺘﻤﺎﻋﻲ ﺧﻼﻝ‬
‫ﺍﻟﻔﺘﺮﺓ ‪) 1992.4 -1988.1‬ﲦﺎﻧﻴﺔ ﻭﲦﺎﻧﻮﻥ ﺍﻟﻔﺼﻞ ﺍﻷﻭﻝ‪-‬ﺍﺛﻨﺎﻥ ﻭﺗﺴﻌﻮﻥ ﺍﻟﻔﺼﻞ‬
‫ﺍﻟﺮﺍﺑﻊ(‪ ،‬ﺍﳌﻄﻠﻮﺏ ﻛﺸﻒ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺑﺎﺳﺘﺨﺪﺍﻡ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ‪.‬‬
‫ﺍﺟﺘﻤﺎﻋﻲ )‪(1992.4- 1988.1‬‬ ‫ﺍﳉﺪﻭﻝ )‪ :(10‬ﻣﻌﻄﻴﺎﺕ ﻓﺼﻠﻴﺔ ﺣﻮﻝ ﻣﺘﻐﲑ‬

‫‪48‬‬
 ‫ﻣﺸﺎﻫﺪﺓ‬ ‫ﺳﻨﺔ‪/‬ﻓﺼﻞ‬ ‫ﻣﺸﺎﻫﺪﺓ‬ ‫ﺳﻨﺔ‪/‬ﻓﺼﻞ‬ ‫ﻣﺸﺎﻫﺪﺓ‬ ‫ﺳﻨﺔ‪/‬ﻓﺼﻞ‬
‫‪60‬‬ ‫‪1991.3‬‬ ‫‪32‬‬ ‫‪1989.4‬‬ ‫‪14‬‬ ‫‪1988.1‬‬
‫‪36‬‬ ‫‪1991.4‬‬ ‫‪12‬‬ ‫‪1990.1‬‬ ‫‪20‬‬ ‫‪1988.2‬‬
‫‪6‬‬ ‫‪1992.1‬‬ ‫‪12‬‬ ‫‪1990.2‬‬ ‫‪44‬‬ ‫‪1988.3‬‬
‫‪11‬‬ ‫‪1992.2‬‬ ‫‪68‬‬ ‫‪1990.3‬‬ ‫‪21‬‬ ‫‪1988.4‬‬
‫‪64‬‬ ‫‪1992.3‬‬ ‫‪29‬‬ ‫‪1990.4‬‬ ‫‪10‬‬ ‫‪1989.1‬‬
‫‪50‬‬ ‫‪1992.4‬‬ ‫‪7‬‬ ‫‪1991.1‬‬ ‫‪19‬‬ ‫‪1989.2‬‬
‫‪18‬‬ ‫‪1991.2‬‬ ‫‪64‬‬ ‫‪1989.3‬‬

‫ﻳﻔﺘﺮﺽ ﰲ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﳌﻌﺮﻓﺔ ﺍﳌﺴﺒﻘﺔ ﻟﻠﻌﻼﻗﺔ ﺍﻟﱵ ﲡﻤﻊ‬

‫ﺑﲔ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﻗﺒﻞ ﺍﻟﻘﻴﺎﻡ ﺑﻌﻤﻠﻴﺔ ﻛﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﺗﻘﺘﻀﻲ‬
‫ﺿﺮﻭﺭﺓ ﺗﻄﺒﻴﻖ ﺍﻻﺧﺘﺒﺎﺭ‪ ،‬ﺍﻟﺮﺟﻮﻉ ﺇﱃ ﻋﻤﻠﻴﺔ ﺍﻟﻜﺸﻒ ﻋﻦ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‬
‫ﻗﺼﺪ ﺇﺯﺍﻟﺘﻬﺎ ﺇﻥ ﻭﺟﺪﺕ ﻭﻣﻨﻪ ﺗﻔﺎﺩﻱ ﺗﻀﻠﻴﻞ ﺍﻟﻨﺘﻴﺠﺔ‪.‬‬
‫ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﻛﺸﻒ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰒ ﺍﻟﻔﺼﻠﻴﺔ ﻭﻓﻖ ﺍﳌﺮﺍﺣﻞ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(11‬ﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﺘﺼﺎﻋﺪﻱ ﳌﺸﺎﻫﺪﺍﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﺍﻟﺮﺗﺒﺔ‬ ‫ﺍﳌﺸﺎﻫﺪﺓ‬ ‫ﺍﻟﻔﺼﻞ‬ ‫ﺍﻟﺮﺗﺒﺔ‬ ‫ﺍﳌﺸﺎﻫﺪﺓ‬ ‫ﺍﻟﻔﺼﻞ‬
‫‪20‬‬ ‫‪68‬‬ ‫‪3‬‬ ‫‪7‬‬ ‫‪14‬‬ ‫‪1‬‬
‫‪12‬‬ ‫‪29‬‬ ‫‪4‬‬ ‫‪10‬‬ ‫‪20‬‬ ‫‪2‬‬
‫‪2‬‬ ‫‪07‬‬ ‫‪1‬‬ ‫‪15‬‬ ‫‪44‬‬ ‫‪3‬‬
‫‪8‬‬ ‫‪18‬‬ ‫‪2‬‬ ‫‪11‬‬ ‫‪21‬‬ ‫‪4‬‬
‫‪17‬‬ ‫‪60‬‬ ‫‪3‬‬ ‫‪3‬‬ ‫‪10‬‬ ‫‪1‬‬
‫‪14‬‬ ‫‪36‬‬ ‫‪4‬‬ ‫‪9‬‬ ‫‪19‬‬ ‫‪2‬‬
‫‪1‬‬ ‫‪6‬‬ ‫‪1‬‬ ‫‪18‬‬ ‫‪64‬‬ ‫‪3‬‬
‫‪4‬‬ ‫‪11‬‬ ‫‪2‬‬ ‫‪13‬‬ ‫‪32‬‬ ‫‪4‬‬
‫‪18‬‬ ‫‪64‬‬ ‫‪3‬‬ ‫‪5‬‬ ‫‪12‬‬ ‫‪1‬‬
‫‪16‬‬ ‫‪50‬‬ ‫‪4‬‬ ‫‪5‬‬ ‫‪12‬‬ ‫‪2‬‬

‫ﳒﺮﻱ ﺍﺧﺘﺒﺎﺭ ﺩﺍﻧﻴﺎﻝ ﻟﻼﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻋﻠﻰ ﺃﺳﺎﺱ ﺗﺮﺗﻴﺐ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺗﺮﺗﻴﺒﺎﹰ‬
‫ﺗﺼﺎﻋﺪﻳﺎﹰ ﻛﻤﺎ ﻳﺒﻴﻨﻪ ﺍﳉﺪﻭﻝ ﺍﻟﺘﺎﱄ‪:‬‬

‫ﺍﳉﺪﻭﻝ )‪ :( 12‬ﺣﺴﺎﺏ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﻓﺮﻭﻕ ﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﺰﻣﲏ ﻭﺍﻟﺘﺼﺎﻋﺪﻱ‬

‫‪49‬‬
 ‫‪d²‬‬ ‫‪d‬‬ ‫‪R‬‬ ‫‪t‬‬
‫‪36‬‬ ‫‪6‬‬ ‫‪7‬‬ ‫‪1‬‬
‫‪64‬‬ ‫‪8‬‬ ‫‪10‬‬ ‫‪2‬‬
‫‪144‬‬ ‫‪12‬‬ ‫‪15‬‬ ‫‪3‬‬
‫‪49‬‬ ‫‪7‬‬ ‫‪11‬‬ ‫‪4‬‬
‫‪4‬‬ ‫‪-2‬‬ ‫‪3‬‬ ‫‪5‬‬
‫‪9‬‬ ‫‪3‬‬ ‫‪9‬‬ ‫‪6‬‬
‫‪121‬‬ ‫‪11‬‬ ‫‪18‬‬ ‫‪7‬‬
‫‪25‬‬ ‫‪5‬‬ ‫‪13‬‬ ‫‪8‬‬
‫‪12.25‬‬ ‫‪-3.5‬‬ ‫‪5.5‬‬ ‫‪9‬‬
‫‪20.25‬‬ ‫‪-4.5‬‬ ‫‪5.5‬‬ ‫‪10‬‬
‫‪81‬‬ ‫‪9‬‬ ‫‪20‬‬ ‫‪11‬‬
‫‪0‬‬ ‫‪0‬‬ ‫‪12‬‬ ‫‪12‬‬
‫‪121‬‬ ‫‪-11‬‬ ‫‪2‬‬ ‫‪13‬‬
‫‪36‬‬ ‫‪-6‬‬ ‫‪8‬‬ ‫‪14‬‬
‫‪4‬‬ ‫‪2‬‬ ‫‪17‬‬ ‫‪15‬‬
‫‪4‬‬ ‫‪-2‬‬ ‫‪14‬‬ ‫‪16‬‬
‫‪256‬‬ ‫‪-16‬‬ ‫‪1‬‬ ‫‪17‬‬
‫‪196‬‬ ‫‪-14‬‬ ‫‪4‬‬ ‫‪18‬‬
‫‪1‬‬ ‫‪-1‬‬ ‫‪18‬‬ ‫‪19‬‬
‫‪16‬‬ ‫‪-4‬‬ ‫‪16‬‬ ‫‪20‬‬
‫‪1199.5‬‬
‫ﻧﺸﲑ ﺇﱃ ﺃﻧﻪ ﰲ ﺍﳊﺎﻟﺔ ﻫﺬﻩ ﻭﻣﻊ ﻭﺟﻮﺩ ﺭﺗﺐ ﻣﻜﺮﺭﺓ‪ ،‬ﻳﺘﻢ ﺍﺳﺘﺨﺪﺍﻡ ﺑﺪﻝ‬
‫ﺍﻟﺮﺗﺒﺔ‪ ،‬ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﻟﻠﺮﺗﺐ ﺍﳌﻜﺮﺭﺓ ﺑﻌﺪ ﲣﺼﻴﺺ ﻗﻴﻤﺔ ﻭﺍﺣﺪﺓ ﻟﻠﻤﺸﺎﻫﺪﺓ ﻛﻤﺎ‬
‫ﺫﹸﻛﺮ ﺳﺎﺑﻘﺎﹰ‪ .‬ﻧﺴﺠﻞ ﺿﻤﻦ ﺍﳉﺪﻭﻝ ﺃﻋﻼﻩ‪ ،‬ﺗﻜﺮﺍﺭ ﺍﻟﺮﺗﺒﺔ ﺍﳋﺎﻣﺴﺔ ﻣﺮﺗﲔ‪ ،‬ﻭﺇﺫﺍ‬
‫ﺧﺼﺼﺖ ﻟﻜﻞ ﻣﺸﺎﻫﺪﺓ ﺭﺗﺒﺔ ﻓﻘﻂ ﺗﺼﺒﺢ ﻟﺪﻳﻨﺎ ﻣﻦ ﺟﺪﻳﺪ ﺭﺗﺒﺔ ﺧﺎﻣﺴﺔ ﻭﺭﺗﺒﺔ‬
‫ﺳﺎﺩﺳﺔ ﻭﻣﻨﻪ ﻳﻜﻮﻥ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﻟﻠﺮﺗﺒﺘﲔ ﺑﻌﺪ ﺍﻟﺘﺨﺼﻴﺺ ﻫﻮ )‪ ،(5.5‬ﻭﻧﻔﺲ‬
‫ﺍﻟﺸﻲﺀ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﺮﺗﺒﺔ )‪ (18‬ﺍﳌﻜﺮﺭﺓ ﻣﺮﺗﲔ ﺃﻳﻀﺎﹰ ﻭﻳﻜﻮﻥ ﻭﺳﻄﻬﺎ ﻫﻮ )‪.(18.5‬‬
‫ﺑﻌﺪ ﺣﺴﺎﺏ ﻣﻌﺎﻣﻞ ﺍﻻﺭﺗﺒﺎﻁ ‪ rs = 0.09‬ﻳﺘﻢ ﻗﺒﻮﻝ ﺍﻟﻔﺮﺿﻴﺔ ﺍﳌﻌﺪﻭﻣﺔ ﺍﻟﻨﺎﺻﺔ‬
‫ﻋﻠﻰ ﻏﻴﺎﺏ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺣﻴﺚ ‪. r < r = 0 .4451‬‬
‫‪s‬‬ ‫‪a‬‬
‫‪2‬‬

‫ﳝﻜﻦ ﺍﻵﻥ ﺗﻄﺒﻴﻖ ﺍﺧﺘﺒﺎﺭ ﻛﺸﻒ ﺍﻟﻔﺼﻠﻴﺔ ﻣﺒﺘﺪﺋﲔ ﺑﻌﻤﻠﻴﺔ ﺇﻋﺎﺩﺓ ﺗﻨﻈﻴﻢ ﻫﺬﻩ‬
‫ﺍﻟﺮﺗﺐ ﻓﺼﻠﻴﺎﹰ ﺃﻭ ﺣﺴﺐ ﺩﺭﺟﺔ ﺍﻟﺪﻭﺭﻳﺔ‪،‬ﻛﻤﺎ ﻫﻮ ﻣﻮﺿﺢ ﰲ ﺍﳉﺪﻭﻝ ﺍﻟﺘﺎﱄ‪:‬‬

‫ﺍﳉﺪﻭﻝ )‪ :(13‬ﺍﻟﺘﺼﻨﻴﻒ ﺍﳉﺪﻭﱄ ﻟﻠﺮﺗﺐ‬

‫‪50‬‬
 ‫ﺍ‪‬ﻤﻮﻉ ‪Ri‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫ﺍﻟﻔﺼﻞ‪ /‬ﺍﻟﺴﻨﺔ‬
‫‪18.5‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫‪5.5‬‬ ‫‪3‬‬ ‫‪7‬‬ ‫‪1‬‬
‫‪36.5‬‬ ‫‪4‬‬ ‫‪8‬‬ ‫‪5.5‬‬ ‫‪9‬‬ ‫‪10‬‬ ‫‪2‬‬
‫‪89‬‬ ‫‪18.5‬‬ ‫‪17‬‬ ‫‪20‬‬ ‫‪18.5‬‬ ‫‪15‬‬ ‫‪3‬‬
‫‪66‬‬ ‫‪16‬‬ ‫‪14‬‬ ‫‪12‬‬ ‫‪13‬‬ ‫‪11‬‬ ‫‪4‬‬

‫‪، p=4‬‬ ‫ﻭﺑﺘﻄﺒﻴﻖ ﺍﻟﻌﻼﻗﺔ ﺍﳌﺬﻛﻮﺭﺓ ﺳﺎﺑﻘﺎﹰ ﺃﻳﻦ‬

‫‪12‬‬ ‫‪p‬‬
‫‪R 2i‬‬
‫= ‪KW‬‬ ‫) ‪å - 3( T + 1‬‬
‫‪T( T + 1 ) i =1 n i‬‬

‫ﺣﻴﺚ ‪ n i = 5‬ﰲ ﻫﺬﺍ ﺍﳌﺜﺎﻝ ﻭ ﻣﻦ ﺍﳉﺪﻭﻝ ﺃﻋﻼﻩ ﻛﻮﻥ ﻛﻞ ﻓﺼﻞ ﺗﻘﺎﺑﻠﻪ‬

‫)‪ (5‬ﺭﺗﺐ‪ ،‬ﻭﻣﻨﻪ ‪ n 1 = n 2 = n 3 = n 4 = 5‬ﺑﻴﻨﻤﺎ )‪ (T‬ﲤﺜﻞ ﺩﺍﺋـﻤﺎﹰ ﻋﺪﺩ‬
‫ﺍﳌﺸﺎﻫـﺪﺍﺕ ﺍﳌﺴﺎﻭﻳﺔ ﻟـ )‪ (20‬ﰲ ﻫﺬﺍ ﺍﳌﺜﺎﻝ‪.‬‬
‫‪12‬‬ ‫‪é R12 R 22 R32 R 42 ù‬‬
‫= ‪KW‬‬ ‫‪ê‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫)‪ú - 3( 20 + 1‬‬
‫‪20 ( 20 + 1) ë n1‬‬ ‫‪n2‬‬ ‫‪n3‬‬ ‫‪n4 û‬‬
‫ﻭﺑﺎﻟﺘﻌﻮﻳﺾ‬
‫‪12‬‬ ‫‪é18 .5‬‬
‫‪2‬‬
‫‪36 .5‬‬ ‫‪2‬‬
‫‪89‬‬ ‫‪66 ù‬‬
‫‪2‬‬ ‫‪2‬‬

‫= ‪KW‬‬ ‫‪ê‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫‪+‬‬ ‫)‪ú - 3( 20 + 1‬‬

‫‪20 ( 20 + 1) ëê 5‬‬ ‫‪5‬‬ ‫‪5‬‬ ‫‪5 úû‬‬

‫‪KW= 16. 46‬‬

‫ﻟﻠﺒﺤﺚ ﻋﻦ ﺍﻟﻔﺼﻠﻴﺔ ﳚﺐ ﻣﻘﺎﺭﻧﺔ ﺍﻹﺣﺼﺎﺀﺓ ﺍﶈﺴﻮﺑﺔ ﺃﻋﻼﻩ ﻣﻊ ﺗﻠﻚ‬

‫ﺍ‪‬ﺪﻭﻟﺔ ﺑـ ‪ 5%‬ﻣﺴﺘﻮﻯ ﻣﻌﻨﻮﻳﺔ ﻭﻣﻌﻄﺎﺓ ﺏ ‪ ، c 23 ,5 % = 7 . 815‬ﻭﻛﻮﻥ ﺍﻷﻭﱃ‬
‫ﺃﻛﱪ ﻣﻦ ﺍ‪‬ﺪﻭﻟﺔ ﻓﺈﻧﻨﺎ ﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ‪ - H 0‬ﺍﻟﱵ ﺗﻨﺺ ﻋﻠﻰ ﻋﺪﻡ ﻭﺟﻮﺩ‬
‫ﺍﻟﻔﺼﻠﻴﺔ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ -‬ﻭﻧﻘﺮ‪ ‬ﺑﻮﺟﻮﺩﻫﺎ‪ ،‬ﳑﺎ ﻳﻘﺘﻀﻲ ﺃﺧﺬﻫﺎ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ ﰲ‬
‫ﻋﻤﻠﻴﺎﺕ ﺍﻟﻨﻤﺬﺟﺔ ﻭﺍﻟﺘﻨﺒﺆ‪.‬‬
‫ﺍﺟﺘﻤﺎﻋﻲ )‪(1992.4- 1988.1‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(9‬ﻣﻌﻄﻴﺎﺕ ﻓﺼﻠﻴﺔ ﺣﻮﻝ ﻣﺘﻐﲑ‬

‫‪51‬‬
 ‫‪.1992.4- 1988.1‬‬ ‫ﺍﳌﺼﺪﺭ‪ :‬ﺑﻠﺪﻳﺔ ﻣﺮﺍﺩ‪ ،‬ﻭﻻﻳﺔ ﺗﻴﺒﺎﺯﺓ‪ ،‬ﺍﳊﺎﻟﺔ ﺍﳌﺪﻧﻴﺔ‪ ،‬ﻣﻌﻄﻴﺎﺕ ﻓﺼﻠﻴﺔ‬

‫ﻧﻔﺲ ﺍﻟﻨﺘﻴﺠﺔ ﺍﻹﺣﺼﺎﺋﻴﺔ ﻳﻌﻜﺴﻬﺎ ﻫﺬﺍ ﺍﻟﺸﻜﻞ ﺍﻟﺒﻴﺎﱐ‪ ،‬ﺣﻴﺚ ﺗﺘﺒﲔ ﺍﳌﺮﻛﺒﺔ‬
‫ﺍﻟﻔﺼﻠﻴﺔ ﺑﺸﻜﻞ ﻭﺍﺿﺢ ﻣﻦ ﺧﻼﻝ ﺍﻟﻘﻤﻢ ﻭﺍﻟﻨﺘﻮﺀﺍﺕ ﺍﳌﻨﺘﻈﻤﺔ‪ .‬ﻛﻤﺎ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﺗﺘﺬﺑﺬﺏ ﺣﻮﻝ ﻭﺳﻂ ﺣﺴﺎﰊ ﺛﺎﺑﺖ ﻭﻫﻮ ﺩﻟﻴﻞ ﻋﻠﻰ ﻏﻴﺎﺏ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫ﻭﻻﺳﺘﺨﺪﺍﻡ ﺍﳌﻘﻴﺎﺱ ﺍﻟﺴﺎﺑﻖ‪ ،‬ﻭﺇﺿﺎﻓﺔ ﻟﺸﺮﻁ ﺗﻮﻓﺮ ﻋﺪﺩ ﻻ ﺑﺄﺱ ﺑﻪ ﻣﻦ‬
‫ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﻟﺬﻱ ﻳﻜﻮﻥ ﻋﻠﻰ ﺍﻷﻗﻞ ﻳﻌﺎﺩﻝ ) ‪ ( n i ´ p‬ﳚﺐ ﻣﺴﺒﻘﺎ ﻣﻌﺮﻓﺔ ﻧﻮﻉ‬
‫ﺍﻟﻌﻼﻗﺔ ﺍﻟﱵ ﺗﺮﺑﻂ ﺑﲔ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻛﺄﻥ ﺗﻜﻮﻥ ﲡﻤﻴﻌﻴﺔ ﺃﻭ ﺟﺪﺍﺋﻴﺔ‪7‬‬

‫ﺃﻱ‪:‬‬
‫‪Y=L*C*S*I‬‬ ‫)‪(2.21‬‬ ‫ﺍ ‪ -‬ﺍﳊﺎﻟﺔ ﺍﳉﺪﺍﺋﻴﺔ‬
‫‪Y= L + C + S + I‬‬ ‫)‪(2.12‬‬ ‫ﺏ‪ -‬ﺍﳊﺎﻟﺔ ﺍﻟﺘﺠﻤﻴﻌﻴﺔ‬
‫ﺫﻟﻚ ﻷﻥ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻻ ﳛﺘﻤﻞ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﺍﻟﺰﻣﻨﻴﺔ ﺃﺛﻨﺎﺀ ﺍﳊﺴﺎﺏ‪ ،‬ﻟﺬﺍ ﻭﺟﺐ ﺇﺯﺍﻟﺘﻬﺎ ﻗﺒﻞ ﺍﻻﺧﺘﺒﺎﺭ‪ ،‬ﻭﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ ﺗﺸﺘﺮﻁ‬
‫ﺍﳌﻌﺮﻓﺔ ﺍﳌﺴﺒﻘﺔ ﳍﺬﻩ ﺍﻟﻌﻼﻗﺔ‪ ،‬ﻭﳌﻌﺮﻓﺘﻬﺎ ﳝﻜﻦ ﺍﻋﺘﻤﺎﺩ ﺃﺳﻠﻮﺑﲔ‪:‬‬

‫‪ - 7‬ﻛﻤﺎ ﳝﻜﻦ ﺃﻥ ﺗﺄﺧﺬ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻷﺷﻜﺎﻝ ﺍﳌﺨﺘﻠﻄﺔ‪.‬‬

‫‪52‬‬
‫‪ -1‬ﺍﻷﺳﻠﻮﺏ ﺍﻟﺒﻴﺎﱐ‪ (Graph inspection) :‬ﺗﻜﻮﻥ ﻭﻓﻖ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ‪،‬‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺫﺍﺕ ﻋﻨﺎﺻﺮ ﲡﻤﻴﻌﻴﺔ ﳌﺎ ﺗﻨﺤﺼﺮ ﺫﺑﺬﺑﺎ‪‬ﺎ ﺑﲔ ﺧﻄﲔ ﻣﺘﻮﺍﺯﻳﲔ‪،‬‬
‫ﺑﻴﻨﻤﺎ ﺗﻜﻮﻥ ﺫﺑﺬﺑﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳉﺪﺍﺋﻴﺔ ﻏﲑ ﺛﺎﺑﺘﺔ ﺍﻟﺸﺪﺓ )ﺗﺒﺎﻳﻦ ﻣﺘﺰﺍﻳﺪ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ‬
‫ﺗﻘﻊ ﺑﲔ ﺧﻄﲔ ﻣﻨﻔﺮﺟﲔ ﻛﻤﺎ ﻳﻈﻬﺮ ﻗﻲ ﺍﻟﺸﻜﻠﲔ ﺍﻟﺘﺎﻟﻴﲔ‪:‬‬
‫اﻟﺸﻜﻞ )‪ :(10‬اﻟﺤﺎﻟﺔ اﻟﺠﺪاﺋﯿﺔ )ب(‬ ‫اﻟﺸﻜﻞ )‪ :(10‬اﻟﺤﺎﻟﺔ اﻟﺘﺠﻤﯿﻌﯿﺔ )ا(‬

‫‪ -2‬ﺍﻷﺳﻠﻮﺏ ﺍﻻﳓﺪﺍﺭﻱ‪ :‬ﻭﻫﻮ ﻣﺒﲏ ﻋﻠﻰ ﺃﺳﺎﺱ ﺗﻘﺪﻳﺮ ﺍﳌﻌﺎﺩﻟﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫_‬
‫‪si = a + b y i‬‬

‫‪i =1,2,...,m‬‬
‫ﻭ ‪ m‬ﳝﺜﻞ ﻋﺪﺩ ﺍﻟﺴﻨﻮﺍﺕ‪.‬‬
‫_‬
‫‪1 p‬‬
‫= ‪yi‬‬ ‫‪å y ij‬‬ ‫ﺃﻳﻦ‬
‫‪p j= 1‬‬

‫‪1 p‬‬ ‫_‬

‫= ‪si‬‬ ‫) ‪å ( y ij - y i‬‬
‫‪2‬‬

‫‪p j=1‬‬

‫‪j = 1, .. .. , p‬‬ ‫ﻭ‬

‫‪53‬‬
 ‫ﳝﻜﻦ ﺗﻘﺪﻳﺮ ﺍﳌﻌﻠﻤﺔ ﺍﳌﻬﻤﺔ ﺑﺎﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﰲ ﺍﻟﻌﻼﻗﺔ ﺃﻋﻼﻩ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪m‬‬ ‫_‬ ‫= _‬
‫^‬ ‫‪å s i yi - m s y‬‬ ‫)‪(2.23‬‬
‫=‪b‬‬ ‫‪i =1‬‬
‫‪m‬‬ ‫‪-‬‬ ‫=‬
‫‪2‬‬
‫‪å yi - m y‬‬ ‫‪2‬‬
‫‪i =1‬‬

‫ﺍﻟﻘﺮﺍﺭ‪ :‬ﺗﻜﻮﻥ ﺍﻟﺴﻠﺴﻠﺔ ﳐﺘﻠﻄﺔ‪ ،‬ﲡﻤﻴﻌﻴﺔ ﺃﻭ ﺟﺪﺍﺋﻴﺔ ﳌﺎ ﻳﻜﻮﻥ ﻭﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪:‬‬

‫^‬
‫‪0. 05 £ b £ 0. 10‬‬
‫^‬
‫‪b < 0. 05‬‬
‫^‬
‫‪b > 0. 10‬‬

‫ﻣﺜﺎﻝ ‪ :11‬ﻟﺪﻳﻨﺎ ﻧﻔﺲ ﻣﻌﻄﻴﺎﺕ ﺍﻻﺳﺘﻬﻼﻙ ﺍﻟﺴﺎﺑﻘﺔ‪ ،‬ﻭﻧﺮﻳﺪ ﻣﻌﺮﻓﺔ ﺷﻜﻞ ﺍﻟﻌﻼﻗﺔ‬
‫ﺍﳌﻮﺟﻮﺩﺓ ﺑﲔ ﻣﺮﻛﺒﺎ‪‬ﺎ‪ ،‬ﺑﻐﻴﺔ ﺇﺯﺍﻟﺘﻬﺎ ﺃﻭ ﳕﺬﺟﺘﻬﺎ‪.‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(14‬ﺍﻟﻌﻼﻗﺔ ﺑﲔ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫‪-‬‬ ‫ﺍﻟﻔﺼل‬
‫‪si‬‬ ‫‪yi‬‬ ‫‪.4‬‬ ‫‪.3‬‬ ‫‪.2‬‬ ‫‪.1‬‬
‫ﺍﻟﺴﻨﺔ‬
‫‪6.057‬‬ ‫‪161.75‬‬ ‫‪171‬‬ ‫‪163‬‬ ‫‪158‬‬ ‫‪155‬‬ ‫‪1‬‬
‫‪7.2586‬‬ ‫‪160.75‬‬ ‫‪172‬‬ ‫‪162‬‬ ‫‪156‬‬ ‫‪153‬‬ ‫‪2‬‬
‫‪7.5828‬‬ ‫‪170‬‬ ‫‪181‬‬ ‫‪173‬‬ ‫‪164‬‬ ‫‪162‬‬ ‫‪3‬‬

‫ﻛﻮﻥ‪ ،m = 3 , p = 4 :‬ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ‪ ،‬ﻭﺑﻐﺾ ﺍﻟﻨﻈﺮ ﻋﻦ ﻋﺪﺩ‬

‫ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﻟﺴﻨﻮﻳﺔ ﺍﶈﺪﻭﺩ ﻭﻟﻐﺮﺽ ﺍﻟﺘﺒﺴﻴﻂ‪ ،‬ﻭﺑﺘﻄﺒﻴﻖ ﺍﻟﻌﻼﻗﺔ )‪ ،(2.23‬ﻗﺪﺭﺕ‬
‫^‬
‫ﻣﻌﻠﻤﺔ ﺍﻻﳓﺪﺍﺭ ﺏ ‪ . b = 0. 09‬ﺣﻴﺚ ﺗﻜﻮﻥ ﺍﻟﻌﻼﻗﺔ ﺑﲔ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ‬
‫^‬
‫ﺍﻟﺰﻣﻨﻴﺔ ﳐﺘﻠﻄﺔ ﻛﻮﻥ ‪.0 . 05 £ b £ 0 .1‬‬

‫‪54‬‬
 ‫ﺏ‪ -‬ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﻏﲑ ﺍﳊﺮﺓ‪:‬‬
‫ﳝﻜﻦ ﺗﻠﺨﻴﺼﻬﺎ ﰲ ﻃﺮﻳﻘﺘﲔ ﺑﺴﻴﻄﲔ ﺗﻌﺘﻤﺪ ﺍﻷﻭﱃ ﻋﻠﻰ ﺗﻘﻨﻴﺔ ﺍﻻﳓﺪﺍﺭ‬
‫ﻭﺍﻟﺜﺎﻧﻴﺔ ﻋﻠﻰ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‪.‬‬
‫ﺏ‪ -1-‬ﺍﻟﻄﺮﻳﻘﺔ ﺍﻻﳓﺪﺍﺭﻳﺔ‪ :‬ﺗﺘﻤﺜﻞ ﰲ ﺍﻓﺘﺮﺍﺽ ﻭﺟﻮﺩ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﰲ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺑـ ‪ p‬ﻣﻦ ﺍﳌﺆﺷﺮﺍﺕ‪ ،‬ﻭﺍﻟﱵ ﻳﻌﱪ ﻋﻨﻬﺎ ﺑـﻨﻔﺲ ﺍﻟﻌﺪﺩ ﻣﻦ‬
‫ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﺘﻤﺜﻴﻠﻴﺔ ﺍﻟﱵ ﻳﺘﻢ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎ‪‬ﺎ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﰒ ﺍﺧﺘﺒﺎﺭﻫﺎ‬
‫ﺇﺣﺼﺎﺋﻴﺎﹰ‪ .‬ﲤﺜﻞ ﺗﻠﻚ ﺍﳌﻌﻠﻤﺎﺕ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ ﺍﳋﺎﺻﺔ ﺑﺎﻟﻈﺎﻫﺮﺓ ﺍﳌﺪﺭﻭﺳﺔ‪.‬‬
‫ﺏ‪ -2-‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‪ :‬ﺗﻌﺘﻤﺪ ﻋﻠﻰ ﻓﻜﺮﺓ ﺍﻻﺭﺗﺒﺎﻁ ﺑﲔ ﺍﳌﺸﺎﻫﺪﺍﺕ‬
‫ﻭﰲ ﻓﺘﺮﺍﺕ ﳐﺘﻠﻔﺔ‪ ،‬ﻛﻤﺎ ﻳﺘﺒﲔ ﰲ ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ ﻣﻦ ﻫﺬﺍ ﺍﻟﻜﺘﺎﺏ‪ ،‬ﺑﺄﻥ ﳍﺬﻩ ﺍﻟﺪﺍﻟﺔ‬
‫ﺃﳘﻴﺔ ﺑﺎﻟﻐﺔ ﰲ ﺇﺑﺮﺍﺯ ﺍﳋﺼﺎﺋﺺ ﺍﳉﻮﻫﺮﻳﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ .‬ﺗﻈﻬﺮ ﺍﻟﻔﺼﻠﻴﺔ ﰲ‬
‫ﺍﻟﺮﺳﻢ ﺍﻟﺒﻴﺎﱐ ﳍﺬﻩ ﺍﻟﺪﺍﻟﺔ ﰲ ﺷﻜﻞ ﻗﻤﻢ ﻭﺍﳔﻔﺎﺿﺎﺕ ﰲ ﻓﺘﺮﺍﺕ ﺯﻣﻨﻴﺔ ﺗﻌﺎﺩﻝ ‪، p‬‬
‫‪.8‬‬
‫ﺃﻱ ﺃﻧﻪ ﺗﻈﻬﺮ ﻗﻤﺔ ﰲ ﺩﻭﺭﺓ ﺗﻌﺎﺩﻝ ‪ p‬ﻭﻧﻔﺲ ﺍﻟﺸﻲﺀ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻼﳔﻔﺎﺿﺎﺕ‬
‫ﻳﺘﻢ ﺣﺴﺎﺏ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﻤﺘﻐﲑ ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ ﺃﻳﻦ‬
‫‪ . k =1, K‬ﰲ ﻫﺬﺍ ﺍﻟﺸﺄﻥ‪ ،‬ﻳﻔﻀﻞ ﺍﻹﺣﺼﺎﺋﻴﻮﻥ ﻭﺑﺸﻜﻞ ﻣﺘﻔﻖ ﻋﻠﻴﻪ ﲢﺪﻳﺪ )‪(K‬‬
‫ﻋﻠﻰ ﺃﺳﺎﺱ ﺃ‪‬ﺎ ﲤﺜﻞ ﺭﺑﻊ ﺣﺠﻢ ﺍﻟﻌﻴﻨﺔ ﺃﻱ ‪: K = T / 4‬‬
‫‪T‬‬

‫‪å(y‬‬
‫‪t =1‬‬
‫‪t‬‬ ‫) ‪- y )( y t - k - y‬‬
‫)‪(2.24‬‬
‫= ‪rk‬‬ ‫‪T‬‬

‫‪å(y‬‬
‫‪t =1‬‬
‫‪t‬‬ ‫‪- y) 2‬‬

‫)‪( 5‬‬‫ﳝﺜﻞ ﺍﻟﺸﻜﻞ)‪ (11‬ﺃﺩﻧﺎﻩ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﻤﺘﻐﲑ ﺍﻟﻮﺍﺭﺩ ﰲ ﺍﳌﺜﺎﻝ‬
‫ﺻﻔﺤﺔ )‪ ،(32‬ﺣﻴﺚ ﺗﻈﻬﺮ ﻗﻤﺔ ﺑﻌﺪ ﻛﻞ ﺃﺭﺑﻌﺔ ﻓﺘﺮﺍﺕ )‪ (12 ،8 ،4‬ﻭﻧﻔﺲ ﺍﻟﺸﻲﺀ‬
‫ﺑﺎﻟﻨﺴﺒﺔ ﻟﻼﳔﻔﺎﺿﺎﺕ ﺍﻟﱵ ﺗﻨﻌﻜﺲ ﰲ ﺍﻟﻔﺘﺮﺍﺕ )‪.(10 ،6 ،2‬ﺃﻣﺎ ﺍﻟﺸﻜﻞ )‪ (12‬ﻻ‬

‫‪ - 8‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ "ﲢﻠﻴﻞ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ"‪ ،‬ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ‪.‬‬

‫‪55‬‬
‫ﻳﻌﻜﺲ ﺃﻱ ﻭﺟﻮﺩ ﻟﻠﺘﺄﺛﲑ ﺍﻟﻔﺼﻠﻲ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﳌﺴﺘﺨﺪﻣﺔ ﰲ ﺣﺴﺎﺏ ﺩﺍﻟﺔ‬
‫ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺗﻠﻚ‪.‬‬

‫ﺑ‪. K = 12‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(11‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﻤﺘﻐﲑ ﺍﻻﺟﺘﻤﺎﻋﻲ‬

‫ﺑ‪K = 15‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(12‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﳌﺘﻐﲑ ﺑﺪﻭﻥ ﻣﺮﻛﺒﺔ ﻓﺼﻠﻴﺔ‬

‫ﻣﺜﺎﻝ ‪ :12‬ﻟﺘﻮﺿﻴﺢ ﻛﻴﻔﻴﺔ ﺣﺴﺎﺏ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻧﺪﺭﺝ ﻫﺬﺍ‬
‫ﺍﳌﺜﺎﻝ ﺍﳌﺘﻌﻠﻖ ﺑﻌﺪﺩ ﺳﻜﺎﻥ ﺍﳉﺰﺍﺋﺮ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪ 2007-1980‬ﻭﺍﻟﻮﺍﺭﺩ ﰲ ﺍﳉﺪﻭﻝ‬
‫)‪ (15‬ﺍﻟﺘﺎﱄ‪ .‬ﻧﻔﺘﺮﺽ ﺃﻥ ‪ ،K=7‬ﻭﻣﻨﻪ ﳓﺴﺐ ﺍﳌﻌﺎﻣﻼﺕ ﺍﻟﺴﺒﻊ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﻟﻌﻼﻗﺔ‬
‫)‪(2.24‬ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪56‬‬
 ‫ﺍﳉﺪﻭﻝ )‪ :(15‬ﺍﳊﺴﺎﺏ ﺍﻟﻴﺪﻭﻱ ﳌﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬
‫‪t‬‬ ‫‪yt‬‬ ‫‪yt - y‬‬ ‫(‬ ‫‪yt - y )²‬‬ ‫‪y t -1 - y‬‬ ‫)‪(6‬‬ ‫)‪(7‬‬
‫‪1‬‬ ‫‪18.66‬‬ ‫‪-8.14‬‬ ‫‪66.19‬‬ ‫…‬ ‫‪0‬‬
‫‪2‬‬ ‫‪19.24‬‬ ‫‪-7.56‬‬ ‫‪57.09‬‬ ‫‪-8.14‬‬ ‫‪61.47‬‬ ‫…‬ ‫‪0‬‬
‫‪3‬‬ ‫‪19.86‬‬ ‫‪-6.94‬‬ ‫‪48.10‬‬ ‫‪-7.56‬‬ ‫‪52.40‬‬ ‫…‬ ‫‪0‬‬
‫‪4‬‬ ‫‪20.51‬‬ ‫‪-6.29‬‬ ‫‪39.51‬‬ ‫‪-6.94‬‬ ‫‪43.60‬‬ ‫‪0‬‬
‫‪5‬‬ ‫‪21.17‬‬ ‫‪-5.63‬‬ ‫‪31.65‬‬ ‫‪-6.29‬‬ ‫‪35.36‬‬ ‫‪0‬‬
‫‪6‬‬ ‫‪22.2‬‬ ‫‪-4.60‬‬ ‫‪21.12‬‬ ‫‪-5.63‬‬ ‫‪25.85‬‬ ‫‪0‬‬
‫‪7‬‬ ‫‪22.8‬‬ ‫‪-4.00‬‬ ‫‪15.97‬‬ ‫‪-4.60‬‬ ‫‪18.36‬‬ ‫‪0‬‬
‫‪8‬‬ ‫‪23.4‬‬ ‫‪-3.40‬‬ ‫‪11.53‬‬ ‫‪-4.00‬‬ ‫‪13.57‬‬ ‫‪32.51‬‬
‫‪9‬‬ ‫‪24.1‬‬ ‫‪-2.70‬‬ ‫‪7.27‬‬ ‫‪-3.40‬‬ ‫‪9.15‬‬ ‫‪25.66‬‬
‫‪10‬‬ ‫‪24.7‬‬ ‫‪-2.10‬‬ ‫‪4.39‬‬ ‫‪-2.70‬‬ ‫‪5.65‬‬ ‫‪18.70‬‬
‫‪11‬‬ ‫‪25.02‬‬ ‫‪-1.78‬‬ ‫‪3.15‬‬ ‫‪-2.10‬‬ ‫‪3.72‬‬ ‫‪13.17‬‬
‫‪12‬‬ ‫‪25.64‬‬ ‫‪-1.16‬‬ ‫‪1.34‬‬ ‫‪-1.78‬‬ ‫‪2.05‬‬ ‫‪9.99‬‬
‫‪13‬‬ ‫‪26.27‬‬ ‫‪-0.53‬‬ ‫‪0.28‬‬ ‫‪-1.16‬‬ ‫‪0.61‬‬ ‫‪5.31‬‬
‫‪14‬‬ ‫‪26.89‬‬ ‫‪0.09‬‬ ‫‪0.01‬‬ ‫‪-0.53‬‬ ‫‪-0.05‬‬ ‫‪2.10‬‬
‫‪15‬‬ ‫‪27.49‬‬ ‫‪0.69‬‬ ‫‪0.48‬‬ ‫‪0.09‬‬ ‫‪0.07‬‬ ‫‪-0.32‬‬
‫‪16‬‬ ‫‪28.06‬‬ ‫‪1.26‬‬ ‫‪1.60‬‬ ‫‪0.69‬‬ ‫‪0.88‬‬ ‫‪-1.87‬‬
‫‪17‬‬ ‫‪28.56‬‬ ‫‪1.76‬‬ ‫‪3.11‬‬ ‫‪1.26‬‬ ‫‪2.23‬‬ ‫‪-2.65‬‬
‫‪18‬‬ ‫‪29.04‬‬ ‫‪2.24‬‬ ‫‪5.04‬‬ ‫‪1.76‬‬ ‫‪3.96‬‬ ‫‪-3.13‬‬
‫‪19‬‬ ‫‪29.5‬‬ ‫‪2.70‬‬ ‫‪7.31‬‬ ‫‪2.24‬‬ ‫‪6.07‬‬ ‫‪-2.59‬‬
‫‪20‬‬ ‫‪29.96‬‬ ‫‪3.16‬‬ ‫‪10.01‬‬ ‫‪2.70‬‬ ‫‪8.56‬‬ ‫‪-1.42‬‬
‫‪21‬‬ ‫‪30.41‬‬ ‫‪3.61‬‬ ‫‪13.06‬‬ ‫‪3.16‬‬ ‫‪11.44‬‬ ‫‪0.30‬‬
‫‪22‬‬ ‫‪30.87‬‬ ‫‪4.07‬‬ ‫‪16.60‬‬ ‫‪3.61‬‬ ‫‪14.73‬‬ ‫‪2.51‬‬
‫‪23‬‬ ‫‪31.35‬‬ ‫‪4.55‬‬ ‫‪20.74‬‬ ‫‪4.07‬‬ ‫‪18.56‬‬ ‫‪5.15‬‬
‫‪24‬‬ ‫‪31.84‬‬ ‫‪5.04‬‬ ‫‪25.44‬‬ ‫‪4.55‬‬ ‫‪22.97‬‬ ‫‪8.04‬‬
‫‪25‬‬ ‫‪32.36‬‬ ‫‪5.56‬‬ ‫‪30.96‬‬ ‫‪5.04‬‬ ‫‪28.07‬‬ ‫‪11.32‬‬
‫‪26‬‬ ‫‪32.9‬‬ ‫‪6.10‬‬ ‫‪37.26‬‬ ‫‪5.56‬‬ ‫‪33.97‬‬ ‫‪15.05‬‬
‫‪27‬‬ ‫‪33.49‬‬ ‫‪6.69‬‬ ‫‪44.81‬‬ ‫‪6.10‬‬ ‫‪40.86‬‬ ‫‪19.32‬‬
‫‪28‬‬ ‫‪33.99‬‬ ‫‪7.19‬‬ ‫‪51.76‬‬ ‫‪6.69‬‬ ‫‪48.16‬‬ ‫‪24.20‬‬
‫‪å‬‬ ‫‪750.28‬‬ ‫‪575.79‬‬ ‫‪-7.19‬‬ ‫‪512.26‬‬ ‫‪181.32‬‬

‫‪57‬‬
 ‫‪28‬‬

‫‪å(y‬‬
‫‪t =1‬‬
‫‪t‬‬ ‫) ‪- y )( y t - k - y‬‬
‫= ‪rk‬‬ ‫‪28‬‬

‫‪å(y‬‬
‫‪t =1‬‬
‫‪t‬‬ ‫‪- y) 2‬‬

‫‪1 28‬‬
‫=‪y‬‬ ‫‪å yt‬‬
‫‪T t =1‬‬
‫ﺣﻴﺚ‬

‫ﺍﻟﺬﻱ ﻳﻌﱪ ﻋﻠﻰ ﺍﳌﺘﻮﺳﻂ ﺍﳊﺴﺎﰊ ﻟﻠﺴﻠﺴﻠﺔ ﺍﳌﺪﺭﻭﺳﺔ ﻭ ﺍﻟﺬﻱ ﻗﺪﺭ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ‬
‫)‪ (7‬ﳝﺜﻼﻥ( ‪yt - y )( y t -1 - y‬‬
‫(‬ ‫ﺑﻴﻨﻤﺎ ﺍﻟﻌﻤﻮﺩﻳﻦ )‪ (6‬ﻭ‬ ‫‪،‬‬ ‫‪y = 26.80‬‬ ‫ﺏ‬
‫)‪ ( yt-7 - y)( yt -1 - y‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪.‬‬
‫ﻭﻓﻖ ﺍﳉﺪﻭﻝ ﺍﳌﻮﺳﻊ ﺍﻟﺴﺎﺑﻖ ﳝﻜﻦ ﺣﺴﺎﺏ ﻣﻌﺎﻣﻼﺕ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺴﺒﻌﺔﻛﻤﺎ‬
‫ﻳﺄﰐ‪:‬‬
‫‪28‬‬

‫‪å (y‬‬
‫‪t =1‬‬
‫‪t‬‬ ‫) ‪- y ) ( y t -1 - y‬‬
‫‪512 . 26‬‬
‫= ‪r1‬‬ ‫‪28‬‬
‫=‬ ‫‪= 0 . 89‬‬
‫‪575 . 79‬‬
‫‪å (y‬‬
‫‪t =1‬‬
‫‪t‬‬ ‫‪- y)2‬‬

‫‪k=K‬‬ ‫ﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ ﻳﺘﻢ ﺣﺴﺎﺏ ﺑﺎﻗﻲ ﺍﳌﻌﺎﻣﻼﺕ ﺍﻻﺧﺮﻯ ﺇﱃ ﻏﺎﻳﺔ‬

‫ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪28‬‬

‫‪å‬‬ ‫) ‪( y t - y )( y t- 2 - y‬‬
‫‪464 . 10‬‬
‫= ‪r2‬‬ ‫‪t =1‬‬
‫‪28‬‬
‫=‬ ‫‪= 0 . 81‬‬
‫‪575 . 79‬‬
‫‪å‬‬‫‪t =1‬‬
‫)‪(yt - y‬‬ ‫‪2‬‬

‫ﺍﻟﱵ ﻳﺘﻢ ﺗﺒﻴﺎ‪‬ﺎ ﰲ ﺍﳉﺪﻭﻝ )‪ (16‬ﺍﻟﺘﺎﱄ‪:‬‬

‫ﺍﳉﺪﻭﻝ )‪ :(16‬ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬
‫‪k‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫‪3‬‬ ‫‪4‬‬ ‫‪5‬‬ ‫‪6‬‬ ‫‪7‬‬
‫‪r‬‬ ‫‪0.89‬‬ ‫‪0.81‬‬ ‫‪0.70‬‬ ‫‪0.60‬‬ ‫‪0.50‬‬ ‫‪0.40‬‬ ‫‪0.31‬‬

‫‪58‬‬
‫ﺣﻴﺚ ﰎ ﺣﺴﺎﺏ ﻫﺬﻩ ﺍﳌﻌﺎﻣﻼﺕ ﺑﻨﻔﺲ ﺍﻟﻌﻼﻗﺔ ﻭﺍﻟﻄﺮﻳﻘﺔ ﺃﻋﻼﻩ‪ .‬ﳔﺘﻢ ﻫﺬﺍ‬
‫ﺍﳌﺜﺎﻝ ﲝﺴﺎﺏ ﺍﳌﻌﺎﻣﻞ ﺍﻟﺴﺎﺑﻊ ﻭﺍﻻﺧﲑ ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪28‬‬

‫‪å (y‬‬ ‫‪t‬‬ ‫) ‪- y ) ( y t -7 - y‬‬

‫‪181.32‬‬
‫= ‪r7‬‬ ‫‪t =1‬‬
‫‪28‬‬
‫=‬ ‫‪= 0.31‬‬
‫‪575.79‬‬
‫‪å(y‬‬
‫‪t =1‬‬
‫‪t‬‬ ‫‪- y) 2‬‬

‫ﺍﻟﱵ ﳝﻜﻦ ﲤﺜﻴﻠﻬﺎ ﺑﻴﺎﻧﻴﺎﹰ ﻟﺘﻌﻜﺲ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻮﺍﺭﺩﺓ ﰲ ﺍﻟﺸﻜﻞ‬
‫ﺍﻟﺘﺎﱄ‪:‬‬
‫ﺍﳉﺰﺍﺋﺮ) ‪(2007- 1980‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(13‬ﺩﺍﻟﺔ ﺍﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﳌﺘﻐﲑ ﺳﻜﺎﻥ‬

‫ﻳﺴﺘﻄﻴﻊ ﺍﻟﻄﺎﻟﺐ ﺑﺎﻟﺘﺠﺮﺑﺔ ﻭﻛﺬﺍ ﺍﳋﱪﺓ ﺍﻟﺘﺒﻴ‪‬ﻦ ﻣﻦ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﺪﺭﻭﺳﺔ‬

‫ﺧﺎﻟﻴﺔ ﻣﻦ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻭﻻ ﺗﺴﲑﻫﺎ ﺇﻻ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺑﺎﻹﺿﺎﻓﺔ‬
‫ﻟﻠﻌﺸﻮﺍﺋﻴﺔ‪.‬‬

‫‪59‬‬
60
 ‫ﺍﻟﻔﺼﻞ ﺍﻟﺜﺎﻟﺚ‬

‫ﺍﻟﺘﻨﺒﺆ ﺑﻨﻤﺎﺫﺝ ﺍﻻﺳﺘﻘﻄﺎﺏ ﺍﻟﺒﺴﻴﻄﺔ‬

‫‪ -1‬ﺍﻟﺘﻨﺒﺆ ﺑﻨﻤﺎﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﻃﺮﻕ ﺗﻘﻴ‪‬ﻴﻤﻬﺎ‬

‫‪‬ﺘﻢ ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﺑﺎﳌﺮﻛﺒﺔ ﺍﻟﻨﻈﺎﻣﻴﺔ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﻭﺍﳌﺘﻤﺜﻠﺔ ﰲ ﺷﻜﻞ ﺍﲡﺎﻩ ﻋﺎﻡ ﺍﻟﺬﻱ ﻗﺪ ﻳﻜﻮﻥ ﻣﺘﻤﺜﻼﹰ ﰲ ﺩﺍﻟﺔ ﺧﻄﻴﺔ‪ ،‬ﺃﺳﻴﺔ ﺃﻭ‬
‫ﻟﻮﻏﺎﺭﻳﺘﻤﻴﺔ ‪...‬ﺍﱁ‪ ،‬ﺇﺿﺎﻓﺔ ﺇﱃ ﻣﺮﻛﺒﺔ ﻋﺸﻮﺍﺋﻴﺔ ﺿﻌﻴﻔﺔ ﺍﻟﺬﺑﺬﺑﺔ‪ .‬ﺗﻔﺴﺮ ﻫﺬﻩ‬
‫ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺘﻐﲑ ﺍﳌﺮﺍﺩ ﺩﺭﺍﺳﺘﻪ ﺑﻮﺍﺳﻄﺔ ﺍﻟﺰﻣﻦ )‪ (t‬ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﺳﺎﺑﻘﺎ‪ ،‬ﻣﻨﻬﺎ‪:‬‬

‫ﺍﳋﻄﻲ ‪Linear Trend Model‬‬ ‫‪ -1-1‬ﳕﻮﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‬

‫ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﱵ ﺗﻨﻤﻮ ﲟﻘﺪﺍﺭ ﻣﻄﻠﻖ ﺛﺎﺑﺖ ﻋﱪ ﺍﻟﺰﻣﻦ‬
‫ﺑﺎﻟﻌﻼﻗﺔ ﺃﻭ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺘﺎﱄ‪:‬‬
‫) ‪yt = f ( t , ut‬‬

‫ﻣﻦ ﺧﻼﻝ ﺍﻟﺘﻤﺜﻴﻞ ﺍﻟﺒﻴﺎﱐ ﻟﻠﺴﻠﺴﻠﺔ ﺍﳌﺪﺭﻭﺳﺔ ﻣﻊ ﺍﻟﺰﻣﻦ‪ ،‬ﻧﺴﺘﻨﺘﺞ ﻭﻛﻤﺎ‬

‫ﺫﻛﺮﻧﺎ ﺳﺎﺑﻘﺎﹰ ﺷﻜﻞ ﺍﻟﺪﺍﻟﺔ )‪ ، f (.‬ﻛﻤﺎ ﻫﻮ ﺍﳊﺎﻝ ﻫﻨﺎ ﻋﻨﺪ ﲤﺜﻴﻞ ﻣﻌﻄﻴﺎﺕ ﺍﳉﺪﻭﻝ‬
‫)‪ (2‬ﺻﻔﺤﺔ )‪ (31‬ﻭﺍﳌﺘﻌﻠﻘﺔ ﺑﺘﻌﺪﺍﺩ ﺳﻜﺎﻥ ﺍﳉﺰﺍﺋﺮ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪،2007 – 1980‬‬
‫ﺃﻳﻦ ﺗﱪﺯ ﺑﻜﻞ ﻭﺿﻮﺡ ﺍﻟﻌﻼﻗﺔ ﺍﳋﻄﻴﺔ )‪ (3.1‬اﻟﺘﺎﻟﯿﺔ ‪:‬‬
‫‪y t = a + b. t + ut‬‬ ‫)‪(3.1‬‬

‫ﻳﻌﻜﺲ ﺍﳌﻨﺤﲎ ﺍﻟﺘﺎﱄ ﺳﻴﻄﺮﺓ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﺇﺿﺎﻓﺔ ﺇﱃ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﺍﻟﻀﻌﻴﻔﺔ ﺍﻟﺬﺑﺬﺑﺔ‪ ،‬ﳑﺎ ﻳﺴﻬﻞ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻮﻗﻊ ﺍﳌﺴﺘﻘﺒﻠﻲ‪.‬‬

‫‪61‬‬
 ‫‪.2007 – 1980‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(14‬ﻋﺪﺩ ﺳﻜﺎﻥ ﺍﳉﺰﺍﺋﺮ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ‬

‫@ @‬
‫‪.2008‬‬ ‫ﺍﳌﺼﺪﺭ‪ :‬ﺍﻟﺒﻨﻚ ﺍﻟﻌﺎﳌﻲ‬

‫ﺣﻴﺚ ‪ t‬ﻫﻨﺎ ﲤﺜﻞ ﺩﻟﻴﻞ ﺍﻟﺰﻣﻦ‪ ،‬ﺃﻱ ﻭﺣﺪﺓ ﻗﻴﺎﺱ ﺍﻟﺴﻨﺔ‪ ،‬ﺍﻟﻔﺼﻞ‪ ،‬ﺍﻟﺸﻬﺮ‬
‫ﻭﻫﻜﺬﺍ‪ ،‬ﻳﺄﺧﺬ ﺍﻟﻘﻴﻢ ﻣﻦ )‪ (1‬ﻛﺴﻨﺔ ﺃﺳﺎﺱ ﰒ ‪ T ، ....3 ،2‬ﻭﻳﻜﻮﻥ ﻫﺬﺍ ﺍﻟﺮﻗﻢ‬
‫ﺍﻷﺧﲑ ﻫﻮ ﻧﻔﺴﻪ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﳌﺴﺘﻌﻤﻠﺔ ﰲ ﺍﻟﺪﺭﺍﺳﺔ ﺃﻱ ﺍﻟﻌﻴﻨﺔ‪ .‬ﺑﻴﻨﻤﺎ ‪b ، a‬‬
‫ﻣﻌﻠﻤﲔ ﻳﺮﺍﺩ ﺗﻘﺪﻳﺮﳘﺎ ﻷﻏﺮﺍﺽ ﺍﻟﺘﻨﺒﺆ ﰲ ﺍﳌﺴﺘﻘﺒﻞ ﺍﻟﻘﺮﻳﺐ‪.‬‬
‫ﻛﻮﻥ ﺍﻟﻌﻼﻗﺔ ﺃﻋﻼﻩ ﺧﻄﻴﺔ ﺍﳌﻌﻠﻤﺎﺕ ﻓﻴﻤﻜﻦ ﺗﻘﺪﻳﺮﻫﺎ ﺑﺎﻟﻄﺮﻳﻘﺔ ﺍﳋﻄﻴﺔ‬
‫ﺍﳌﺸﻬﻮﺭﺓ ﻭﺍﳌﻌﺮﻭﻓﺔ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﺍﻟﻌﺎﺩﻳﺔ ‪(Ordinary Least‬‬
‫)‪.Squares‬‬

‫‪ -1‬ﻃﺮﻳﻘﺔ ﺍﳌﻌﺎﺩﻻﺕ ﺍﻟﻄﺒﻴﻌﻴﺔ‬

‫ﳝﻜﻦ ﺍﳊﺼﻮﻝ ﻋﻠﻰ ﻫﺬﻩ ﺍﳌﻘﺪﺭﺍﺕ ﺇﺫﺍ ﺍﻋﺘﱪﻧﺎ ﺍﳌﻌﺎﺩﻟﺔ ﺃﺩﻧﺎﻩ ﻛﻨﻈﲑﺓ ﻟﻠﻌﻼﻗﺔ‬
‫ﺃﻋﻼﻩ ﺃﻱ‪:‬‬
‫^‬ ‫^‬
‫‪y t = a + b. t + e t‬‬
‫ﺑـﻴﻨﻤﺎ ‪ et‬ﲤﺜﻞ ﺗﻘﺪﻳﺮ ﺍﻷﺧﻄﺎﺀ ﻭﺍﻟﱵ‬ ‫‪b,a‬‬ ‫ﻫﻲ ﻣﻘﺪﺭﺍﺕ‬ ‫ﺣﻴﺚ‬
‫ˆ‪bˆ , a‬‬
‫ﻧﺴﻤﻴﻬﺎ ﺍﻟﺒﻮﺍﻗﻲ )‪.(Residuals‬‬

‫‪62‬‬
‫ﺗﻌﺘﻤﺪ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﰲ ﺍﻟﺘﻘﺪﻳﺮ ﻋﻠﻰ ﻣﺒﺪﺃ ﺗﺼﻐﲑ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‬
‫ﺍﻟﺬﻱ ﻳﺮﻣﺰ ﺇﻟﻴﻪ ﺑـ ‪ RSS‬ﺃﻱ‪:‬‬
‫‪T‬‬ ‫‪T‬‬ ‫^‬ ‫^‬
‫‪Min å e 2t = å ( y t - a - b. t ) 2‬‬
‫‪t =1‬‬ ‫‪t =1‬‬
‫‪T‬‬
‫ﻭﻛﻮﻥ ‪ Min å e 2t‬ﺗﺘﻮﺍﻓﻖ ﻣﻊ ﻧﻘﻄﺔ ﺍﻧﻌﻄﺎﻑ ﺻﻐﺮﻯ ﺃﻳﻦ ﺗﻜﻮﻥ ﺍﳌﺸﺘﻘﺔ‬
‫‪t =1‬‬
‫ﺍﻷﻭﱃ ﳍﺎ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﻌﻠﻤﺘﲔ ﻣﻌﺪﻭﻣﺔ ﺣﻴﺚ‪:‬‬
‫‪T‬‬
‫‪¶ å et2‬‬ ‫‪T‬‬ ‫^‬ ‫^‬
‫‪t =1‬‬
‫^‬
‫‪= -2å ( yt - a - b .t ) = 0‬‬
‫‪¶a‬‬ ‫‪i =1‬‬

‫‪T‬‬ ‫^‬ ‫‪^ T‬‬

‫‪= å yt - T a- b å t = 0‬‬
‫‪t =1‬‬ ‫‪t =1‬‬

‫ﻭﻫﻲ ﺍﳌﻌﺎﺩﻟﺔ ﺍﻟﻄﺒﻴﻌﻴﺔ ﺍﻷﻭﱃ‪.‬‬

‫‪T‬‬ ‫^‬ ‫^‬ ‫‪T‬‬

‫‪å y t = T a+ bå t‬‬
‫‪t =1‬‬ ‫‪t =1‬‬

‫ﻭﻣﻨﻪ ﻭﺑﻌﻤﻠﻴﺔ ﺑﺴﻴﻄﺔ ﳒﺪ ﺃﻥ‬

‫^‬ ‫_‬ ‫_ ^‬
‫‪a = y- b t‬‬ ‫)‪(3.2‬‬

‫_‬
‫‪1 T‬‬
‫=‪y‬‬ ‫‪åy t‬‬
‫‪T t =1‬‬
‫ﺣﻴﺚ‬

‫_‬
‫‪1 T‬‬ ‫‪T+1‬‬
‫=‪t‬‬ ‫‪å‬‬ ‫=‪t‬‬ ‫ﻭ‬
‫‪T t =1‬‬ ‫‪2‬‬

‫‪63‬‬
 ‫ﻭﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ‬
‫‪T‬‬
‫‪¶ å et2‬‬ ‫‪T‬‬ ‫^‬ ‫^‬
‫‪t =1‬‬
‫^‬
‫‪= -2å ( yt - a - b .t )t = 0‬‬
‫‪¶b‬‬ ‫‪t =1‬‬
‫‪T‬‬ ‫^‬ ‫‪T‬‬ ‫^‬ ‫‪T‬‬
‫‪= å y t t - aå t - bå t 2 = 0‬‬
‫‪t =1‬‬ ‫‪t =1‬‬ ‫‪t =1‬‬
‫‪T‬‬ ‫^‬ ‫‪T‬‬ ‫^‬ ‫‪T‬‬

‫‪åy‬‬ ‫‪t‬‬ ‫‪t = aå t + bå t 2‬‬ ‫ﻫﻲ ﺍﳌﻌﺎﺩﻟﺔ ﺍﻟﻄﺒﻴﻌﻴﺔ ﺍﻟﺜﺎﻧﻴﺔ‬

‫‪t =1‬‬ ‫‪t =1‬‬ ‫‪t =1‬‬
‫^‬
‫ﺑﺘﻌﻮﻳﺾ ‪ a‬ﺑﻘﻴﻤﺘﻬﺎ ﺍﶈﺴﻮﺑﺔ ﺃﻋﻼﻩ‪:‬‬
‫‪T‬‬ ‫_‬ ‫_ ^‬ ‫‪T‬‬ ‫^‬ ‫‪T‬‬

‫‪å y t t = ( y - b t )å t + bå t 2‬‬
‫‪t =1‬‬ ‫‪t =1‬‬ ‫‪t =1‬‬

‫_‬ ‫‪T‬‬ ‫_ ^‬ ‫‪T‬‬ ‫^‬ ‫‪T‬‬

‫‪= y å t - b t å t + bå t 2‬‬
‫‪t =1‬‬ ‫‪t =1‬‬ ‫‪t =1‬‬

‫ﺑﻀﺮﺏ ﺍﳌﻘﺪﺍﺭﻳﻦ ﺍﻷﻭﻟﲔ ﰲ ﺍﻟﻄﺮﻑ ﺍﻷﳝﻦ ﰲ ‪ T‬ﻭﺍﻟﻘﺴﻤﺔ ﻋﻠﻴﻪ ﳓﺼﻞ‬

‫ﻋﻠﻰ ﻣﻘﺎﺩﻳﺮ ﺟﺪﻳﺪﺓ ﺑﺪﻻﻟﺔ ﺍﻷﻭﺳﺎﻁ ﺍﳊﺴﺎﺑﻴﺔ ﲟﻌﲎ‪:‬‬
‫‪T‬‬ ‫_ _‬ ‫^‬ ‫‪T‬‬ ‫‪_2‬‬

‫) ‪å y t t - T y t = b( å t 2 - T t‬‬
‫‪t =1‬‬ ‫‪t =1‬‬
‫‪T‬‬ ‫‪-‬‬ ‫‪-‬‬
‫‪å ty‬‬ ‫‪t‬‬ ‫‪-T y t‬‬
‫ﻣﻨﻪ‬
‫^‬
‫=‪b‬‬ ‫‪t =1‬‬ ‫)‪(3.3‬‬
‫‪T‬‬ ‫‪-‬‬
‫‪åt‬‬ ‫‪2‬‬
‫‪-T t‬‬ ‫‪2‬‬

‫ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﺃﻳﻀﺎﹰ ﻋﻨﻬﺎ ﺑﺪﻻﻟﺔ ﺍﳓﺮﺍﻑ ﺍﻟﻘﻴ‪‬ﻢ ﻋﻦ ﺃﻭﺳﺎﻃﻬﺎ ﺍﳊﺴﺎﺑﻴﺔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪t =1‬‬

‫‪T‬‬ ‫‪-‬‬ ‫‪-‬‬

‫^‬ ‫‪å ( t - t )( y‬‬ ‫‪t‬‬ ‫)‪-y‬‬

‫=‪b‬‬ ‫‪t =1‬‬ ‫)‪(3.4‬‬
‫‪T‬‬ ‫‪-‬‬
‫) ‪å( t - t‬‬
‫‪t =1‬‬

‫‪64‬‬
‫ﻫﺬﺍ ﺍﻟﺘﻘﺪﻳﺮ ﰎﹼ ﻃﺒﻌﺎﹰ ﲢﺖ ﻓﺮﺿﻴﺎﺕ ﺃﺳﺎﺳﻴﺔ ﻧﺴﻤﻴﻬﺎ ﺑﺎﻟﻔﺮﺿﻴﺎﺕ‬
‫ﺍﻟﻜﻼﺳﻴﻜﻴﺔ ﺗﺴﻤﺢ ﻓﻴﻤﺎ ﺑﻌﺪ ﺑﺎﺷﺘﻘﺎﻕ ﺃﺩﻭﺍﺕ ﺇﺣﺼﺎﺋﻴﺔ ﻣﻔﻴﺪﺓ ﰲ ﻋﻤﻠﻴﺎﺕ‬
‫ﺍﻻﺧﺘﺒﺎﺭ ﺍﳋﺎﺻﺔ ﺑﺘﻘﻴﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ‪.‬‬

‫ﻣﺜﺎﻝ ‪ :13‬ﻟﺘﻮﺿﻴﺢ ﻋﻤﻠﻴﺔ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﱵ ﺍﻟﻨﻤﻮﺫﺝ‪ ،‬ﻧﺴﺘﻌﲔ ﲟﻌﻄﻴﺎﺕ ﺍﳉﺪﻭﻝ ﺭﻗﻢ‬
‫)‪ (2‬ﺍﻟﻮﺍﺭﺩ ﰲ ﺍﻟﺼﻔﺤﺔ )‪ (31‬ﺍﳌﺘﻌﻠﻖ ﺑﺴﻜﺎﻥ ﺍﳉﺰﺍﺋﺮ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪.2007-1980‬‬

‫ﻟﻴﻜﻦ ‪ yt‬ﻣﺘﻐﲑ ﺍﻟﺴﻜﺎﻥ‪ ،‬ﻭﺑﻌﺪ ﺣﺴﺎﺏ ﻣﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﻐﲑﺍﺕ ﺍﳌﺨﺘﻠﻔﺔ‬

‫ﻭﳎﺎﻣﻴﻌﻬﺎ ﻭﻣﺮﺑﻌﺎ‪‬ﺎ ﺍﳌﺘﻤﺜﻠﺔ ﰲ‪:‬‬
‫‪28‬‬ ‫‪28‬‬

‫‪å t = 406‬‬
‫‪1‬‬
‫‪å ty‬‬
‫‪1‬‬
‫‪t‬‬ ‫‪= 11901.94‬‬
‫‪t = 145‬‬
‫‪28‬‬ ‫‪28‬‬
‫‪y = 26.8‬‬
‫‪å‬‬‫‪1‬‬
‫‪y t = 750 .28‬‬ ‫‪åt‬‬
‫‪1‬‬
‫‪2‬‬
‫‪= 7714‬‬

‫ﺑﺘﻄﺒﻴﻖ ﻋﻼﻗﺔ ﺗﻘﺪﻳﺮ ﻣﻴﻞ ﺍﳌﻨﺤﲎ‪:‬‬

‫)‪11901 .94 - 28( 26.8)(14 .5‬‬
‫= ˆ‪b‬‬ ‫‪= 0.56‬‬
‫‪7714 - 28(14 .5)²‬‬
‫ﻣﻌﻄﺎﺓ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬ ‫̂‪a‬‬ ‫ﺗﻜﻮﻥ ﺍﳌﻌﻠﻤﺔ‬
‫‪aˆ = y - bˆt‬‬
‫‪= 18.68‬‬
‫‪‬ﺬﺍ ﺍﻟﺸﻜﻞ ﻳﺼﺒﺢ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻘﺪﺭ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪yˆ t = 18.68 + 0.56t‬‬

‫ﻫﻲ ﻣﻌﺎﺩﻟﺔ ﺍﳋﻂ ﺍﳌﺴﺘﻘﻴﻢ ﻭﺍﳌﻌﱪﺓ ﻋﻠﻰ ﻣﻨﺤﲎ ﻋﺪﺩ ﺳﻜﺎﻥ ﺍﳉﺰﺍﺋﺮ ﻭﺍﻟﱵ‬
‫ﳝﻜﻦ ﺍﺳﺘﺨﺪﺍﻣﻬﺎ ﰲ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﺑﻌﺪ ﺇﺟﺮﺍﺀ ﻋﻤﻠﻴﺔ ﺍﻻﺧﺘﺒﺎﺭ ﺍﻹﺣﺼﺎﺋﻲ ﺍﳌﻮﻓﻖ‬
‫ﻋﻠﻰ ﻣﻌﺎﱂ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻘﺪﺭ‪ .‬ﺳﻨﺴﺘﻌﺮﺽ ﳎﻤﻮﻋﺔ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﻫﺬﻩ ﺿﻤﻦ ﻣﻮﺿﻮﻉ‬
‫ﺗﻘﻴﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﻮﺍﺭﺩ ﰲ ﺍﻟﺼﻔﺤﺔ )‪.(74‬‬

‫‪65‬‬
‫ﻧﺸﲑ ﺇﱃ ﺃﻥ ﻃﺮﻳﻘﺔ ﺍﳌﻌﺎﺩﻻﺕ ﺍﻟﻄﺒﻴﻌﻴﺔ ﺫﺍﺕ ﺍﳌﻌﻠﻤﺘﲔ ﻻ ﺗﺼﻠﺢ ﺳﻮﻯ‬
‫ﳊﺎﻻﺕ ﺍﺣﺘﻮﺍﺀ ﺍﻟﻨﻤﻮﺫﺝ ﻋﻠﻰ ﻣﻌﻠﻤﺘﲔ ﻓﻘﻂ‪ .‬ﺃﻣﺎ ﰲ ﺍﳊﺎﻻﺕ ﺍﻟﻌﺎﻣﺔ ﺍﻷﺧﺮﻯ‬
‫ﳝﻜﻦ ﺍﻟﻠﺠﻮﺀ ﺇﱃ ﻃﺮﻳﻘﺔ ﺍﳌﺼﻔﻮﻓﺎﺕ ﺍﻟﱵ ﺳﻴﺄﰐ ﺫﻛﺮﻫﺎ ﻓﻴﻤﺎ ﻳﻠﻲ‪.‬‬

‫‪ -2‬ﻃﺮﻳﻘﺔ ﺍﳌﺼﻔﻮﻓﺎﺕ‬
‫ﲤﺜﻞ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺗﻌﻤﻴﻤﺎﹰ ﻟﺘﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ ﺑﺪﻭﻥ ﲢﺪﻳﺪ‪ ،‬ﻛﻤﺎ ﺃ‪‬ﺎ‬
‫ﺗﻘﺪﻡ ﺣﻼﹰ ﺭﻳﺎﺿﻴﺎﹶ ﻣﺒﺴﻄﺎﹰ ﻭﳐﺘﺼﺮﺍﹰ ﻟﻨﻤﻮﺫﺝ ﻋﺎﻡ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺧﺼﺎﺋﺺ ﺍﳌﺼﻔﻮﻓﺎﺕ‪.‬‬
‫ﻟﺘﺒﺴﻴﻂ ﺍﻟﻔﻜﺮﺓ‪ ،‬ﻧﺴﺘﻌﲔ ﺑﻨﻔﺲ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺴﺎﺑﻖ ﻭﺗﻌﻮﻳﺾ ‪ t‬ﰲ ﺍﳌﻌﺎﺩﻟﺔ‬
‫ﺃﻋﻼﻩ )‪ (3.1‬ﺑـ ‪ 1‬ﻭ ‪ 2‬ﻭ‪...‬ﻭ‪ T‬ﻟﻨﺤﺼﻞ ﻋﻠﻰ ‪ T‬ﻣﻦ ﺍﳌﻌﺎﺩﻻﺕ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y1 = a + b.1 + u1‬‬
‫‪y 2 = a + b.2 + u 2‬‬
‫‪......‬‬
‫‪yT = a + b.T + uT‬‬
‫ﳝﻜﻦ ﺍﻵﻥ ﺻﻴﺎﻏﺔ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻻﺕ ﰲ ﺷﻜﻞ ﻣﺼﻔﻮﻓﺎﰐ ﻭﻛﺎﻵﰐ‪:‬‬
‫‪é y1 ù é1‬‬ ‫‪1ù‬‬ ‫‪é u1 ù‬‬
‫‪ê y ú ê1‬‬ ‫‪2ú‬‬ ‫‪ú‬‬ ‫‪êu ú‬‬
‫‪ê 2ú ê‬‬ ‫‪éa ù ê ú‬‬
‫‪2‬‬

‫‪ê y3 ú = ê1‬‬ ‫‪3 ú.ê ú + ê u3 ú‬‬ ‫)‪(3.5‬‬

‫‪ê ú ê‬‬ ‫‪ú ëb û ê ú‬‬
‫‪ê : ú ê:‬‬ ‫‪:ú‬‬ ‫‪ê:ú‬‬
‫‪êë yT úû êë1‬‬ ‫‪T úû‬‬ ‫‪êëuT úû‬‬

‫∙‪Y = X‬‬ ‫‪B + U‬‬

‫)‪(T×1) (T×2) (2×1) (T×1‬‬ ‫ﺫﺍﺕ ﺍﻷﺑﻌﺎﺩ‬
‫ﻳﻜﻮﻥ ﻛﻤﺎ ﻳﻠﻰ‪:‬‬ ‫‪b, a‬‬ ‫ﻭﺣﻞ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﻣﻦ ﺃﺟﻞ ﻣﻘﺪﺭﺍﺕ‬
‫^‬ ‫‪é^ ù‬‬
‫‪B = êa^ ú = ( X ' X ) -1 X 'Y‬‬ ‫)‪(3.6‬‬
‫‪êëb úû‬‬

‫‪66‬‬
‫ﺃﻱ‬ (Non singular) ‫ﻏﲑ ﺷﺎﺫﺓ‬ ( X 'X ) ‫ﺑﺸﺮﻁ ﺃﻥ ﺗﻜﻮﻥ ﺍﳌﺼﻔﻮﻓﺔ‬
:‫( ﺃﻳﻦ‬Invertible) ‫ﺫﺍﺕ ﳏﺪﺩ ﻏﲑ ﻣﻌﺪﻭﻡ ﻭﺑﺎﻟﺘﺎﱄ ﻗﺎﺑﻠﺔ ﻟﻠﻘﻠﺐ‬
é1 1 ù
ê1 2 ú
é1 1 . . 1 ù ê ú
X'X = ê ú ê . . ú
ë1 2 . . T û ê ú
ê. . ú
êë1 T úû

:‫ﻭﺑﺘﻄﺒﻴﻖ ﻗﺎﻋﺪﺓ ﺿﺮﺏ ﻣﺼﻔﻮﻓﺘﲔ ﳓﺼﻞ ﻋﻠﻰ‬

éT
ê
å t ùú
=ê ú
êå t å û
t 2ú
ë
‫ﺃﻳﻦ‬
T T( T + 1 )
åt=
t=1 2
T = (T + 1)
2
‫ﺣﻴﺚ‬
T T( T +1 )( 2 T+ 1 )
åt =
2

t =1 6
‫ﻛﺬﻟﻚ‬
é y1 ù
êy ú
é1 1 . . 1 ù ê ú
2

X 'Y = ê
1 2 . . T ú . ú
ê
ë ûê . ú
ê ú
êë yT úû
é å yt ù
ê ú
=ê ú
êå t. yt ú
ë û

67
‫)‪(2‬‬ ‫ﻣﺜﺎﻝ ‪ :14‬ﻟﺘﻮﺿﻴﺢﹴ ﺃﻛﺜﺮ ﻟﻄﺮﻳﻘﺔ ﺍﻟﺘﻘﺪﻳﺮ‪ ،‬ﻧﻌﻴﺪ ﺍﺳﺘﺨﺪﺍﻡ ﻣﻌﻄﻴﺎﺕ ﺍﳉﺪﻭﻝ‬
‫ﻭﺍﳌﻤﺜﻠﺔ ﺑﻴﺎﻧﻴﺎﹰ ﰲ ﺍﻟﺸﻜﻞ )‪ ،(14‬ﻭﺫﻟﻚ ﺑﻐﻴﺔ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﱵ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳋﻄﻲ ﺍﻟﺘﺎﱄ‬
‫ﺍﻟﺬﻱ ﻳﻌﻜﺴﻪ ﺍﻟﺸﻜﻞ ﺍﳌﺬﻛﻮﺭ ﺳﺎﺑﻘﺎﹰ‪.‬‬
‫‪y t = a + b. t + u t‬‬

‫ﺣﻴﺚ ‪ yt‬ﳝﺜﻞ ﻛﻤﺎ ﺳﺒﻖ ﺫﻛﺮﻩ ﻋﺪﺩ ﺳﻜﺎﻥ ﺍﳉﺰﺍﺋﺮ ﺑﻴﻨﻤﺎ ‪ t‬ﻳﻌﺘﱪ ﺩﻟﻴﻼ‬
‫ﻟﻠﺰﻣﻦ‪.‬‬
‫) ‪ ( X ' X‬ﻭ ) ‪( X 'Y‬‬ ‫ﻟﺘﻘﺪﻳﺮ ﻣﻌﻠﻤﱵ ﺍﻟﻨﻤﻮﺫﺝ‪ ،‬ﻧﺸﻜﻞ ﺃﻭﻻﹰ ﺍﳌﺼﻔﻮﻓﺘﲔ‬
‫ﻛﻤﺎ ﻳﻠﻲ ‪:‬‬
‫‪é 28 406 ù‬‬
‫‪(X ' X ) = ê‬‬ ‫‪ú‬‬
‫‪ë406 7714û‬‬
‫ﺑﻴﻨﻤﺎ‬
‫‪é 750 .28 ù‬‬
‫‪( X 'Y ) = ê‬‬ ‫‪ú‬‬
‫‪ë11901 .94 û‬‬
‫ﻧﻌﻴﺪ ﻛﺘﺎﺑﺔ ﻣﻌﺎﺩﻟﺔ ﺍﳊﻞ ﻛﻤﺎ ﻳﻠﻲ ‪:‬‬
‫^‬ ‫‪é^ ù‬‬
‫‪B = êa^ ú = (X ' X )-1 X 'Y‬‬
‫‪êëbúû‬‬
‫ﺣﻴﺚ‬
‫^‬ ‫‪é ù‬‬
‫^‬
‫‪1‬‬ ‫‪é 7714 - 406ù é 750.28 ù‬‬
‫= ‪B = êa^ ú‬‬ ‫‪ê‬‬ ‫‪.‬‬
‫‪êëb úû - 51156 ë- 406‬‬ ‫‪28 úû êë11901.94úû‬‬

‫ﺃﻳﻦ ﻳﻜﻮﻥ ﻣﻘﺪﺍﺭ ﺍﳌﺼﻔﻮﻓﺔ ﺍﳌﻌﻜﻮﺳﺔ ﻣﻌﻄﻰ ﺏ‪:‬‬

‫‪1 é 7714 - 406ù‬‬

‫= ‪( X ' X ) -1‬‬
‫‪- 51156 êë- 406 28 úû‬‬

‫‪68‬‬
 ‫ﺑﻴﻨﻤﺎ ﺍﻟﺮﻗﻢ )‪ (-51156‬ﳝﺜﻞ ﳏﺪﺩ ﺍﳌﺼﻔﻮﻓﺔ ) ‪ ، ( X ' X‬ﻭﻣﻨﻪ‬
‫^‬ ‫‪é ^ ù é18 . 68 ù‬‬
‫‪B = ê a^ ú = ê‬‬ ‫‪ú‬‬
‫‪êë b úû ë 0 . 56 û‬‬

‫ﻧﻌﱪ ﻋﻦ ﻫﺬﻩ ﺍﻷﺭﻗﺎﻡ ﰲ ﺷﻜﻞ ﻣﻌﺎﺩﻟﺔ ﻣﻘﺪﺭﺓ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪yˆ t = 18.68 + 0.56t‬‬
‫ﻳﺘﻢ ﺗﺸﻜﻴﻞ ﺗﻨﺒﺆ ﺍﻟﻔﺘﺮﺓ ﺍﳌﻮﺍﻟﻴﺔ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﻟﻌﻼﻗﺔ ﺍﳌﻘﺪﺭﺓ ﻛﻤﺎ ﻳﻠﻲ‬
‫‪1‬‬
‫‪:‬‬

‫)‪yˆ T +1 = aˆ + bˆ(T + 1‬‬

‫ﻫﻮ ﺗﻨﺒﺆ ‪ Y‬ﰲ ﺍﻟﻔﺘﺮﺓ ‪ T+1‬ﺍﻧﻄﻼﻗﺎ ﻣﻦ ‪.T‬‬ ‫‪$y T+1 /T‬‬ ‫ﺣﻴﺚ‬
‫^‬ ‫^‬ ‫^‬
‫) ‪y T +2 = a + b ( T + 2‬‬
‫ﻭﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ﺗﻜﺘﺐ ﻣﻌﺎﺩﻟﺔ ﺍﻟﺘﻨﺒﺆ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫^‬ ‫^‬ ‫^‬
‫) ‪y T+ L = a + b(T + L‬‬ ‫)‪(3.7‬‬

‫ﺃﻳﻦ ‪ L‬ﳝﺜﻞ ﺃﻓﻖ ﺍﻟﺘﻨﺒﺆ‪ ،‬ﻭﺑﺘﺤﻠﻴﻞ ﺃﻭﺳﻊ ﳍﺬﻩ ﺍﻟﻌﻼﻗﺔ ﻭﺑﻔﻚ ﺍﻷﻗﻮﺍﺱ ﳝﻜﻦ‬
‫ﺍﻟﺘﻮﺻﻞ ﺇﱃ‪:‬‬
‫^‬ ‫^‬ ‫^‬ ‫^‬
‫‪y T+L = (a + b T ) + b L‬‬
‫^‬ ‫^‬ ‫^‬
‫‪yT+L = yT + b L‬‬ ‫)‪(3.8‬‬
‫^‬ ‫^‬ ‫^‬
‫ﰲ ﺍﻟﻌﻼﻗﺔ )‪:(3.7‬‬ ‫‪L =0‬‬ ‫ﳌﺎ‬ ‫) ‪y T = ( a+ b T‬‬ ‫ﺣﻴﺚ‬
‫^‬
‫ﻭﻣﺎﺩﺍﻡ ‪ b‬ﻫﻮ ﻣﻘﺪ‪‬ﺭ ﻣﻘﺪﺍﺭ ﺍﻟﻨﻤﻮ ﺍﳌﻄﻠﻖ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ‪ y t‬ﺃﻱ‪:‬‬
‫^‬ ‫^‬ ‫^‬ ‫^‬ ‫^‬ ‫^‬
‫) ) ‪y T - y T-1 = ( a + b T ) - ( a + b( T - 1‬‬

‫‪ -1‬ﻻ ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﺇﻻ ﺑﻌﺪ ﲡﺎﻭﺯ ﻣﺮﺣﻠﺔ ﺗﻘﻴ‪‬ﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ "‪ "Model validation‬ﺑﺎﺳﺘﺨﺪﺍﻡ‬
‫ﺃﺩﻭﺍﺕ ﺍﻻﺧﺘﺒﺎﺭ ﺍﳌﻨﺎﺳﺒﺔ ] ‪ ،[ ...c2 ,Fisher ,Student ,R 2‬ﺍﳌﻮﻓﺮﺓ ﻣﻦ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻘﺪﻳﺮ‪.‬‬

‫‪69‬‬
 ‫^‬ ‫^‬ ‫^‬
‫‪y T - y T -1 = b‬‬
‫ﻳﺘﻢ ﻛﺘﺎﺑﺔ ﺍﻟﻌﻼﻗﺔ )‪ (3.8‬ﰲ ﺍﻟﺸﻜﻞ‪:‬‬
‫^‬ ‫^‬ ‫^‬ ‫^‬
‫‪y T + L = y T + ( y T - y T -1 ) L‬‬

‫ﻣﺎﺩﺍﻣﺖ ﻗﻴ‪‬ﻢ ﻛﻞ ﻣﻦ ‪ y t‬ﻭ ‪ yt -1‬ﻗﺪ ﲢﻘﻘﺖ‪ ،‬ﻭﻟﺘﺤﺴﲔ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ‪،‬‬

‫ﻧﻌﻮﺽ ﻫﺬﻩ ﺍﻟﻘﻴ‪‬ﻢ ﺑﺘﻠﻚ ﺍﳌﺸﺎﻫﺪﺓ‪:‬‬
‫^‬
‫‪y T + L = y T + ( y T - y T -1 ) L‬‬ ‫)‪(3.9‬‬

‫ﻭﺍﻟﱵ‬ ‫)‪(Holt-Winters‬‬ ‫ﻫﺬﻩ ﻫﻲ ﻧﻘﻄﺔ ﺍﻧﻄﻼﻕ ﻃﺮﻳﻘﺔ ﻫﻮﻟﺖ‪-‬ﻭﻧﺘﺮﺯ‬

‫ﺳﻨﺮﺍﻫﺎ ﰲ ﺍﻟﻔﺼﻞ ﺍﻟﺜﺎﻟﺚ ﺿﻤﻦ ﳎﻤﻮﻋﺔ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﻜﻴﻔﺔ‪.‬‬
‫ﺍﻟﺸﻜﻞ )‪ (15‬ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﻟﻜﻠﻲ‬

‫ﺇﻥ ﺍﻟﺘﻨﺒﺆ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻨﻤﻮﺫﺝ ﺭﻗﻢ )‪ (3.1‬ﻳﻜﻮﻥ ﺻﺎﺋﺒﺎ ﻭﺇﳚﺎﺑﻴﺎ ﰲ ﺣﺎﻟﺔ‬

‫ﻛﻮﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺫﺍﺕ ﺍﲡﺎﻩ ﻋﺎﻡ ﻛﻠﻲ )‪ ،(Global trend‬ﻛﻤﺎ ﻫﻮ ﻣﺒﲔ ﰲ‬
‫ﺍﻟﺸﻜﻞ)‪ ،(15‬ﺑﻴﻨﻤﺎ ﺇﺫﺍ ﻛﺎﻥ ﻫﺬﺍ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﳏﻠﻴ‪‬ﺎ ﻓﻘﻂ )‪– (Local trend‬‬
‫ﺍﻟﺸﻜﻞ )‪ ،(16‬ﺍﳌﻌﱪ ﻋﻦ ﺣﺎﻟﺔ ﺗﻐﲑ ﻫﻴﻜﻠﻲ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﳑﺎ ﻳﻘﺘﻀﻲ‬
‫ﺗﻜﻴﻴﻒ ﻋﻤﻠﻴﺔ ﺗﺼﻤﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻨﺎﺳﺐ ﻟﻴﺄﺧﺬ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ ﺫﻟﻚ ﺍﻟﺘﻐﲑ ﰲ‬

‫‪70‬‬
‫ﻣﺴﺎﺭ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ .‬ﻟﺬﺍ ﻭﰲ ﺣﺎﻟﺔ ﻭﺟﻮﺩ ﺍﲡﺎﻩ ﻋﺎﻡ ﳏﻠﻲ‪ ،‬ﻭﻗﺒﻞ ﺍﺧﺘﻴﺎﺭ ﺍﻟﺘﻘﻨﻴﺔ‬
‫ﺍﳌﻨﺎﺳﺒﺔ ﻟﻠﺘﻄﺒﻴﻖ‪ ،‬ﳚﺐ ﺃﻥ ﺗﺄﺧﺬ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ ﺍﻟﺘﻐﲑ ﰲ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺮﻗﻤﻴﺔ ﻋﻦ‬
‫ﻃﺮﻳﻖ ﺍﻟﺴﻤﺎﺡ ﻟﻠﻤﻌﻠﻤﺎﺕ ﺑﺘﻌﺪﻳﻞ ﻧﻔﺴﻬﺎ ﻟﻜﻞ ﻭﺿﻊ ﺟﺪﻳﺪ‪.‬‬
‫ﺍﻟﺸﻜﻞ )‪ :(16‬ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﶈﻠﻲ‬

‫‪The exponential model‬‬ ‫‪ -2-1‬ﺍﻟﻨﻤﻮﺫﺝ ﺍﻷﺳﻲ‪:‬‬

‫ﻓﺈﺫﺍ ﻛﺎﻧﺖ ‪ y t‬ﺗﻨﻤﻮ ﺑﺎﳌﻘﺪﺍﺭ ﺍﳌﻄﻠﻖ ﺍﻟﺜﺎﺑﺖ ﰲ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺴﺎﺑﻖ‪ ،‬ﻧﻔﺘﺮﺿﻬﺎ‬
‫ﰲ ﻫﺬﺍ ﺍﻷﺧﲑ ﺗﻨﻤﻮ ﺑﻨﺴﺒﺔ ﻣﺌﻮﻳﺔ ﺛﺎﺑﺘﺔ ﻣﻘﺪﺍﺭﻫﺎ )‪ (r‬ﺃﻱ‪:‬‬
‫‪y t = A e rt‬‬ ‫)‪(3.10‬‬

‫ﻧﺸﲑ ﻫﻨﺎ ﺇﱃ ﻭﺟﻮﺩ ﺃﺷﻜﺎﻝ ﻣﺘﻌﺪﺩﺓ ﺗﻘﻊ ﲢﺖ ﻫﺬﺍ ﺍﻻﺳﻢ ﺍﻟﻌﺮﻳﺾ‪،‬‬

‫ﻭﻧﻘﺘﺼﺮ ﻫﻨﺎ ﻋﻠﻰ ﻫﺬﻩ ﺍﻟﺼﻴﻐﺔ ﻓﻘﻂ ﺍﻟﱵ ﲤﻴﺰﺕ ﺑﻌﺪﻡ ﺍﳋﻄﻴﺔ ﰲ ﻛﻞﹺ ﻣﻦ ﺍﳌﻌﻠﻤﺎﺕ‬
‫ﻭﺍﳌﺘﻐﲑﺍﺕ‪ .‬ﻟﺘﺴﻬﻴﻞ ﻋﻤﻠﻴﺔ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ‪ ،‬ﳓﻮﻝ ﺍﳌﻌﺎﺩﻟﺔ ﻋﻦ ﻃﺮﻳﻖ‬
‫ﺍﻟﻠﻮﻏﺎﺭﻳﺘﻢ ﺇﱃ ﺩﺍﻟﺔ ﺧﻄﻴﺔ ﻛﺎﻷﰐ ﺑﻌﺪ ﺇﺿﺎﻓﺔ ﺍﳋﻄﺄ ﺍﻟﻌﺸﻮﺍﺋﻲ ﰲ ﺷﻜﻞ ﺃﺳﻲ‬
‫ﻟﺘﺴﻬﻴﻞ ﻋﻤﻠﻴﺔ ﺍﻟﺘﺤﻮﻳﻞ ﻫﺬﻩ‪.‬‬
‫‪rt + u t‬‬
‫‪yt =A e‬‬
‫‪Ln ( y t ) = Ln ( A ) + r t + u t‬‬ ‫)‪(3.11‬‬

‫‪71‬‬
‫ﺣﻴﺚ ) ‪ Ln ( A‬ﻭ ‪ r‬ﻣﻌﻠﻤﺎﺕ ﳚﺐ ﺗﻘﺪﻳﺮﻫﺎ‪ ،‬ﻭﺑﻨﻔﺲ ﺍﻟﺘﻘﻨﻴﺔ ﺍﻟﺴﺎﺑﻘﺔ ﺃﻱ‬
‫‪ ،OLS‬ﺑﻌﺪ ﺇﻋﺎﺩﺓ ﺗﺴﻤﻴﺔ ﺍﳌﻘﺎﺩﻳﺮ ﺍﻟﻠﻮﻏﺎﺭﲤﻴﺔ ﲟﻘﺎﺩﻳﺮ ﻋﺎﺩﻳﺔ ﺃﻱ‪:‬‬
‫‪Ln ( y t ) = z t‬‬

‫‪Ln ( A ) = a‬‬
‫ﺗﺼﺒﺢ ﺍﳌﻌﺎﺩﻟﺔ ﻣﻦ ﺟﺪﻳﺪ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪z t = a+ rt + u t‬‬
‫ﻳﺘﻢ ﺗﻘﺪﻳﺮ ﺍﳌﻌﻠﻤﺘﲔ ﺍﻷﺧﺮﻳﲔ ﺑﻨﻔﺲ ﺍﳋﻄﻮﺍﺕ ﺍﻟﺴﺎﺑﻘﺔ ﻭﺍﻟﻮﺍﺭﺩﺓ ﰲ ﳕﻮﺫﺝ‬
‫ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﳋﻄﻲ‪.‬‬
‫‪T‬‬ ‫‪-‬‬ ‫‪-‬‬

‫^‬
‫) ‪å ( t - t )( z t - z‬‬
‫=‪r‬‬
‫‪t =1‬‬ ‫)‪(3.12‬‬
‫‪T‬‬ ‫‪-‬‬
‫) ‪å( t - t‬‬
‫‪2‬‬

‫‪t =1‬‬

‫^‬ ‫_‬ ‫_ ^‬
‫‪a = z- r t‬‬ ‫)‪(3.13‬‬ ‫ﻭ‬
‫ﺣﻴﺚ‬
‫_‬
‫‪1‬‬ ‫‪T‬‬
‫= ‪z‬‬ ‫‪å z‬‬
‫‪T t =1 t‬‬
‫ﻛﻤﺎ ﺃﻥ ‪ r‬ﻭ ‪ a‬ﻫﻲ ﻣﻘﺪﺭﺍﺕ ﺛﻮﺍﺑﺖ ﺍﻟﻨﻤﻮﺫﺝ ﺍﶈﻮ‪‬ﻝ‪ ،‬ﻭﻣﻨﻪ ﻓﺎﻟﺘﻨﺒﺆ ﺑـ ‪y‬‬
‫ﰲ ﺍﻟﻔﺘﺮﺓ ‪ l‬ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﳌﺎ ﻳﻜﻮﻥ) ‪ ( l = 1 , L‬ﻫﻮ‪:‬‬
‫^‬
‫^‬ ‫^‬
‫‪y T+1 = A . e r‬‬
‫) ‪( T+ 1‬‬

‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫^‬
‫^‬ ‫^‬
‫‪y T+ L = A . e r‬‬
‫) ‪( T+ L‬‬
‫)‪(3.14‬‬

‫ﳝﻜﻦ ﺍﳊﺼﻮﻝ ﻋﻠﻰ ﻧﻔﺲ ﻫﺬﻩ ﺍﻟﺘﻨﺒﺆﺍﺕ ﻣﺒﺎﺷﺮﺓ‪ ،‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻌﻼﻗﺔ‬

‫ﺍﻟﻠﻮﻏﺎﺭﲤﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫^‬ ‫^‬ ‫^‬
‫)‪zT +1 = a+ r (T + 1‬‬

‫‪72‬‬
 ‫^‬ ‫^‬ ‫^‬
‫)‪zT + L = a+ r(T + L‬‬ ‫)‪(3.15‬‬ ‫ﺑﺎﻟﺘﺸﺎﺑﻪ‬
‫ﻳﺘﻢ ﲢﻮﻳﻞ ﻫﺬﻩ ﺍﻷﺭﻗﺎﻡ ﺇﱃ ﺃﺻﻠﻬﺎ ﻋﻦ ﻃﺮﻳﻖ ﺍﻟﺪﺍﻟﺔ ﺍﻷﺳﻴﺔ ﻭﻛﺎﻵﰐ‪:‬‬
‫^‬ ‫^‬
‫) ‪y T+ 1 = exp( z T +1‬‬
‫^‬
‫ﻭ‬
‫^‬
‫) ‪y T+ L = exp( z T+ L‬‬

‫ﻣﺜﺎﻝ ‪ :15‬ﻟﺪﻳﻨﺎ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ ﺍﳌﻤﺜ‪‬ﻠﺔ ﻟﻠﻨﺎﺗﺞ ﺍﻟﻮﻃﲏ ﺍﻹﲨﺎﱄ ﺑﺎﻷﺭﻗﺎﻡ ﺍﻻﲰﻴﺔ‬
‫ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪ ،91-85‬ﺍﳌﻄﻠﻮﺏ ﺍﻟﺘﻨﺒﺆ ‪‬ﺬﺍ ﺍﳌﺘﻐﲑ ﻟﻠﺴﻨﻮﺍﺕ ‪ 92‬ﻭ ‪ 94‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‬
‫ﺫﻟﻚ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻨﺎﺳﺐ‪.‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(17‬ﺍﻟﻨﺎﺗﺞ ﺍﻟﻘﻮﻣﻲ ﺍﻹﲨﺎﱄ‬
‫‪1991 1990 1989 1988 1987 1986 1985‬‬ ‫ﺍﻟﺴﻨﻮﺍﺕ‬
‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪T‬‬
‫‪735‬‬ ‫‪497‬‬ ‫‪383‬‬ ‫‪317‬‬ ‫‪306‬‬ ‫‪286‬‬ ‫‪287‬‬ ‫‪Y‬‬

‫ﻗﺒﻞ ﺑﺪﺍﻳﺔ ﺃﻱ ﻋﻤﻠﻴﺔ ﺗﻘﺪﻳﺮﻳﺔ‪ ،‬ﳚﺐ ﺍﻟﺘﺄﻛﺪ ﻣﺴﺒﻘﺎ ﻣﻦ ﻣﺮﻛﺒﺎﺕ ﻫﺬﻩ‬

‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﱵ ﻧﺴﻤﻴﻬﺎ ‪ y t‬ﻋﻦ ﻃﺮﻳﻖ ﺍﻟﺘﻤﺜﻴﻞ ﺍﻟﺒﻴﺎﱐ ﻭﺍﻻﺧﺘﺒﺎﺭ ﺍﻹﺣﺼﺎﺋﻲ‪.‬‬
‫ﺍﻟﺸﻜﻞ )‪ :(17‬ﻣﻨﺤﲎ ﺍﻟﻨﺎﺗﺞ ﺍﻟﻘﻮﻣﻲ ﺍﻹﲨﺎﱄ‪.‬‬

‫‪73‬‬
‫ﺑﺪﺍﻳﺔ ﻭﻣﻦ ﺍﳌﻼﺣﻈﺔ ﺍﻟﺮﻗﻤﻴﺔ ﺍﻟﺒﺴﻴﻄﺔ ﻳﺘﺒﲔ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﲡﺎﻩ ﻋﺎﻡ‪،‬‬
‫ﻭﻟﺘﻮﺿﻴﺢ ﻃﺒﻴﻌﺘﻪ‪ ،‬ﻧﺴﺘﻌﲔ ﺑﺎﻟﺸﻜﻞ )‪ (17‬ﺍﻟﺬﻱ ﻳﺆﻛﺪ ﻭﺟﻮﺩﻫﺎ ﻭﻓﻖ ﺍﻟﺸﻜﻞ‬
‫ﺍﻷﺳﻲ‪ ،‬ﻭﺍﻟﱵ ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻨﻬﺎ ﺭﻳﺎﺿﻴ‪‬ﺎ ﺑﻌﺪﺓ ﺃﺷﻜﺎﻝ ﻣﻨﻬﺎ‪:‬‬
‫‪Pib t = y t = Ae rt + u t‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(18‬ﺗﻘﺪﻳﺮ ﺛﻮﺍﺑﺖ ﺍﻟﺪﺍﻟﺔ ﺍﻷﺳﻴﺔ‬
‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫ﺭﻗﻡ ﺍﻟﺨﺎﻨﺔ‬
‫‪4‬‬ ‫‪2‬‬
‫‪5X 4‬‬ ‫‪z -z‬‬ ‫‪t- t‬‬ ‫‪ln( y ) = z‬‬ ‫‪y‬‬ ‫‪t‬‬ ‫ﺍﻟﺴﻨﺔ‬
‫‪9‬‬ ‫‪0.81‬‬ ‫‪-0.27‬‬ ‫‪3-‬‬ ‫‪5.66‬‬ ‫‪287‬‬ ‫‪1‬‬ ‫‪85‬‬
‫‪4‬‬ ‫‪0.56‬‬ ‫‪-0.28‬‬ ‫‪2-‬‬ ‫‪5.65‬‬ ‫‪286‬‬ ‫‪2‬‬ ‫‪86‬‬
‫‪1‬‬ ‫‪0.21‬‬ ‫‪-0.21‬‬ ‫‪1-‬‬ ‫‪5.72‬‬ ‫‪306‬‬ ‫‪3‬‬ ‫‪87‬‬
‫‪0‬‬ ‫‪0‬‬ ‫‪-0.17‬‬ ‫‪0‬‬ ‫‪5.76‬‬ ‫‪317‬‬ ‫‪4‬‬ ‫‪88‬‬
‫‪1‬‬ ‫‪0.20‬‬ ‫‪0.20‬‬ ‫‪1‬‬ ‫‪5.95‬‬ ‫‪383‬‬ ‫‪5‬‬ ‫‪89‬‬
‫‪4‬‬ ‫‪0.54‬‬ ‫‪0.27‬‬ ‫‪2‬‬ ‫‪6.21‬‬ ‫‪497‬‬ ‫‪6‬‬ ‫‪90‬‬
‫‪9‬‬ ‫‪2.01‬‬ ‫‪0.67‬‬ ‫‪3‬‬ ‫‪6.6‬‬ ‫‪735‬‬ ‫‪7‬‬ ‫‪91‬‬
‫‪28‬‬ ‫‪4.15‬‬ ‫‪0‬‬ ‫‪0‬‬ ‫‪41.55‬‬ ‫‪2811‬‬ ‫‪28‬‬ ‫‪å‬‬
‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫ﺍﻟﺨﺎﻨﺔ‬

‫ﻟﺘﻘﺪﻳﺮ ﺛﻮﺍﺑﺖ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺑﺘﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ )‪،(O.L.S‬‬

‫ﳚﺐ ﲢﻮﻳﻞ ﻫﺬﻩ ﺍﻟﻌﻼﻗﺔ ﺇﱃ ﺍﻟﺸﻜﻞ ﺍﳋﻄﻲ ﻋﻦ ﻃﺮﻳﻖ ﺍﻟﻠﻮﻏﺎﺭﻳﺘﻢ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪Ln ( y t ) = Ln ( A ) + r t + u t‬‬
‫‪z t = a+ rt + u t‬‬
‫ﺑﺘﻄﺒﻴﻖ ﺍﻟﻄﺮﻳﻘﺔ ﺍﳌﺬﻛﻮﺭﺓ ﻭﺍﻻﺳﺘﻌﺎﻧﺔ ﺑﺎﳌﻌﻠﻮﻣﺎﺕ ﺍﳌﻘﺪﻣﺔ ﺃﻋﻼﻩ ﰲ ﺍﳉﺪﻭﻝ‬
‫)‪ (18‬ﳓﺼﻞ ﻋﻠﻰ‪:‬‬
‫‪4.15‬‬
‫= ˆ‪r‬‬ ‫‪= 0.15‬‬
‫‪28‬‬
‫‪aˆ = 5.93 - 0.15(4) = 5.33‬‬
‫ﻳﻜﻮﻥ ﺍﻟﺘﻨﺒﺆ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫)‪zˆT +1 = zˆ92 = aˆ + rˆ(T +1‬‬
‫‪= 5. 33 + 0. 15( 8 ) = 6. 53‬‬

‫‪74‬‬
 ‫ﺑﺎﻟﺘﺎﱄ‪:‬‬
‫^‬ ‫^‬
‫) ‪y T + 1 = exp ( z T + 1‬‬
‫^‬
‫) ‪y T + 1 = exp( 6 . 53‬‬

‫‪= 685 . 39‬‬

‫ﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ‪:‬‬
‫^‬ ‫^‬
‫‪y T+ 3 = y 94 = 925. 19‬‬
‫ﺍﳌﻼﺣﻆ ﺍﻗﺘﺼﺎﺩﻳﺎ ﺃﻥ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﻳﻨﺘﻘﺪ ﻋﻠﻰ ﺃﺳﺎﺱ ﺃﻥ ﺍﻟﻈﻮﺍﻫﺮ‬
‫ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻻ ﺗﻨﻤﻮ ﲟﻘﺪﺍﺭ ﻣﻄﻠﻖ ﻋﱪ ﺍﻟﺰﻣﻦ ﺇﱃ ﻣﺎﻻ ‪‬ﺎﻳﺔ‪ ،‬ﺫﻟﻚ ﻟﻮﺟﻮﺩ ﻗﻴﻮﺩ‬
‫ﻛﺜﲑﺓ ﺍﻗﺘﺼﺎﺩﻳﺔ ﺃﻭ ﻏﲑﻫﺎ‪ ،‬ﲢﺪ ﻣﻦ ﻫﺬﺍ ﺍﻟﻨﻤﻮ‪ ،‬ﻛﺤﺪﻭﺙ ﺍﻷﺯﻣﺎﺕ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺃﻭ‬
‫ﺗﺪﺧﻞ ﺍﻟﺪﻭﻟﺔ ﰲ ﻣﺜﺎﻟﻨﺎ ﻫﺬﺍ ﻟﻠﺤﺪ ﻣﻦ ﺍﻟﺘﺪﻫﻮﺭ ﺍﻟﻜﺒﲑ ﰲ ﻗﻴﻤﺔ ﺍﻟﻮﺣﺪﺓ ﺍﻟﻨﻘﺪﻳﺔ ﻣﻦ‬
‫ﺟﻬﺔ‪ ،‬ﺃﻭ ﺍﻟﺘﺤﻜﻢ ﰲ ﻣﻌﺪﻻﺕ ﺍﻟﺘﻀﺨﻢ ﻣﻦ ﺟﻬﺔ ﺛﺎﻧﻴﺔ‪ .‬ﻛﻤﺎ ﺃﻥ ﺑﻌﺾ ﺍﻟﺴﻠﻊ‬
‫ﺍﳉﺪﻳﺪﺓ ﻳﺰﺩﺍﺩ ﺍﻟﻄﻠﺐ ﻋﻠﻴﻬﺎ ﺑﺸﻜﻞ ﻛﺒﲑ ﰒ ﻣﺎ ﻳﻔﺘﺄ ﺃﻥ ﻳﺼﻞ ﺇﱃ ﻣﺴﺘﻮﻯ‬
‫ﺍﻹﺷﺒﺎﻉ‪ ،‬ﺣﻴﺚ ﻳﺘﻮﻗﻒ ﳕﻮ ﺍﻟﻄﻠﺐ ﻋﻠﻰ ﻫﺬﻩ ﺍﻟﺴﻠﻌﺔ‪ ،‬ﻭﻫﻮ ﻣﺎ ﻧﺘﻌﺮﺽ ﻟﻪ ﰲ‬
‫ﺍﻟﺪﺍﻟﺔ ﺍﻟﻠﻮﺟﺴﺘﻴﺔ‪.‬‬
‫‪The quadratic trend model‬‬ ‫‪ -3-1‬ﺩﺍﻟﺔ ﺍﻟﻘﻄﻊ ﺍﳌﻜﺎﻓﺊ‪:‬‬
‫ﻣﻦ ﺑﲔ ﺍﻟﺪﻭﺍﻝ ﺍﻟﻘﺮﻳﺒﺔ ﻟﻠﺸﻜﻞ ﺍﳋﻄﻲ ﳒﺪ ﺩﺍﻟﺔ ﺍﻟﻘﻄﻊ ﺍﳌﻜﺎﻓﺊ ﺍﻟﱵ ﺗﻌﺘﱪ‬
‫ﺍﻣﺘﺪﺍﺩ ﻟﻨﻤﻮﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﳋﻄﻲ ﺍﳌﺴﺘﻌﻤﻞ ﰲ ﲢﻠﻴﻞ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﺍﻟﺬﻱ‬
‫ﻳﻜﺘﺐ ﻋﻠﻲ ﺍﻟﺸﻜﻞ‪:‬‬
‫‪y t = b0 + b1 t + b2 t 2 + u t‬‬ ‫)‪(3.16‬‬

‫ﺣﻴﺚ ﻳﺘﻤﻴﺰ ﺑﺄﻧﻪ ﻏﲑ ﺧﻄﻲ ﺍﳌﺘﻐﲑﺍﺕ‪ .‬ﻫﺬﺍ ﺍﻟﺸﻜﻞ ﻻ ﻳﻄﺮﺡ ﻣﺸﻜﻠﺔ ﰲ‬

‫ﺍﻟﺘﻘﺪﻳﺮ‪ ،‬ﺣﻴﺚ ﻧﻌﻴﺪ ﺗﺴﻤﻴﺔ ‪ t 2‬ﺑـ ‪ x t‬ﻭﻧﻌﻴﺪ ﻛﺘﺎﺑﺔ ﺍﳌﻌﺎﺩﻟﺔ ﻣﻦ ﺟﺪﻳﺪ ﰲ‬
‫ﺍﻟﺸﻜﻞ ﺍﳌﻤﺎﺛﻞ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪y t = b 0 + b1 t + b 2 x t + u t‬‬ ‫)‪(3.17‬‬

‫‪75‬‬
‫ﺗﻌﱪ ﻫﺬﻩ ﺍﻟﺼﻴﺎﻏﺔ ﺍﻟﺮﻳﺎﺿﻴﺔ ﻋﻦ ﳕﻮﺫﺝ ﺧﻄﻲ ﻋﺎﻡ ﺍﻟﺬﻱ ﳝﻜﻦ ﺗﻘﺪﻳﺮ‬
‫ﻣﻌﻠﻤﺎﺗﻪ ﺍﻟﺜﻼﺛﺔ ﺑﻮﺍﺳﻄﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﻭ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺼﻔﻮﻓﺎﺕ ﻭ ﻛﻤﺎ ﻳﻠﻲ ‪:‬‬
‫‪é b^ ù‬‬ ‫‪-1‬‬

‫^‬
‫‪ê ^0 ú é T‬‬
‫‪ê‬‬
‫‪åt‬‬ ‫‪åx‬‬ ‫‪t‬‬ ‫‪ù‬‬
‫‪ú‬‬
‫‪é å yt ù‬‬
‫‪ê‬‬ ‫‪ú‬‬ ‫)‪(3.18‬‬
‫‪B = ê b1 ú = ê å t‬‬ ‫‪åt‬‬ ‫‪2‬‬
‫‪åt x‬‬ ‫‪t ú‬‬ ‫‪ê å t yt ú‬‬
‫‪ê ^ ú‬‬
‫‪ê b 2 ú êë å x t‬‬ ‫‪åt x‬‬ ‫‪t‬‬ ‫‪åx‬‬ ‫‪2‬‬
‫‪t û‬‬
‫‪ú‬‬ ‫‪êå xt y t ú‬‬
‫‪ë‬‬ ‫‪û‬‬
‫‪ëê ûú‬‬

‫ﺃﻱ ﻳﻜﺘﺐ ﺑﺎﻟﺸﻜﻞ ﺍﳌﺨﺘﺼﺮ‪:‬‬

‫^‬
‫‪B = ( X ' X ) -1 X ' Y‬‬
‫)‪( K ´ 1) = ( K ´ K ).(K ´ 1‬‬

‫ﺗﺒﻘﻰ ﻫﺬﻩ ﺍﻟﻘﺎﻋﺪﺓ ﺻﺎﳊﺔ ﻟﺘﻘﺪﻳﺮ ﺃﻱ ﻋﺪﺩ ﻣﻦ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﺎﺫﺝ‬

‫ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﻟﻘﻴﺎﺳﻴﺔ ﻭﺍﻟﱵ ﻧﺸﲑ ﳍﺎ ﺑﺎﳊﺮﻑ )‪ (K‬ﺣﻴﺚ ﺗﺄﺧﺬ ﺍﻟﻌﻼﻗﺔ ﺍﻟﻌﺎﻣﺔ ﺃﻋﻼﻩ‬
‫ﺍﻷﺑﻌﺎﺩ ﺍﳌﺬﻛﻮﺭﺓ ﺃﻋﻼﻩ‪.‬‬

‫‪The logistic curve model‬‬

‫‪2‬‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﻠﻮﺟﺴﱵ‪:‬‬ ‫‪4- 1‬‬

‫ﺗﺴﺘﻌﻤﻞ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﻟﺘﻔﺴﲑ ﺍﻟﻈﻮﺍﻫﺮ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﻟﱵ ﻳﺘﻮﻗﻒ ﳕﻮﻫﺎ ﺑﻌﺪ ﻓﺘﺮﺓ‬
‫ﺯﻣﻨﻴﺔ ﻣﻌﻴﻨﺔ ﻭ ﻫﺬﻩ ﺇﺣﺪﻯ ﺍﻟﺼﻴ‪‬ﻎ ﺍﻟﱵ ﺗﻜﺘﺐ ‪‬ﺎ‪:‬‬
‫‪1‬‬ ‫)‪(3.19‬‬
‫= ‪yt‬‬
‫‪b 0 + b 1 b 2t‬‬

‫ﺣﻴﺚ ‪ b b , b‬ﻣﻌﻠﻤﺎﺕ ﳚﺐ ﺗﻘﺪﻳﺮﻫﺎ ﺑﻄﺮﻳﻘﺔ ﻣﻼﺋﻤﺔ‪ ،‬ﻭﺍﻟﱵ ﺗﻌﻄﻲ‬ ‫‪2 ,‬‬ ‫‪1‬‬ ‫‪0‬‬

‫ﺍﳌﻨﺤﲎ ﺃﺩﻧﺎﻩ ﺇﺫﺍ ﻛﺎﻧﺖ ‪ b > 0‬ﻭ ‪ 0 < b < 1‬ﻭﺃﻳﻦ ‪ 1‬ﲤﺜﻞ ﻣﺴﺘﻮﻱ ﺍﻹﺷﺒﺎﻉ‬
‫‪b0‬‬
‫‪2‬‬ ‫‪1‬‬

‫)‪.(Saturation level‬‬

‫ﻟـ‪.Pearl-Reed‬‬ ‫‪ -2‬ﻣﻨﺤﲎ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﻟﻠﻮﺟﺴﱵ‬

‫‪76‬‬
 ‫ﺍﻟﺸﻜﻞ )‪ :(18‬ﺍﳌﻨﺤﲎ ﺍﻟﻠﻮﺟﺴﱵ‬

‫@ @‬

‫ﺗﺘﻤﻴﺰ ﺍﻟﺪﺍﻟﺔ ﺍﻟﻠﻮﺟﺴﺘﻴﺔ ﻋﻦ ﺳﺎﺑﻘﺎ‪‬ﺎ ﰲ ﻛﻮ‪‬ﺎ ﻏﲑ ﺧﻄﻴﺔ ﺍﳌﻌﻠﻤﺎﺕ ﻭﻏﲑ‬

‫ﻗﺎﺑﻠﺔ ﻟﻠﺘﺤﻮﻳﻞ ﺇﱃ ﺷﻜﻞ ﺧﻄﻲ ﻭﻓﻖ ﺍﻷﺳﻠﻮﺏ ﺍﳌﺸﺎﺭ ﺇﻟﻴﻪ ﺳﺎﺑﻘﺎﹰ‪ .‬ﳍﺬﺍ ﻳﺘﻢ‬
‫ﺗﻘﺪﻳﺮﻫﺎ ﺑﺎﻟﻄﺮﻕ ﻏﲑ ﺍﳋﻄﻴﺔ‪.‬‬
‫ﺍﻗﺘﺒﺴﺖ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﻭﺑﺴﺒﺐ ﺍﻻﺳﺘﻌﻤﺎﻝ ﺍﻟﺸﺎﺋﻊ ﳍﺎ ﰲ ﻋﻠﻢ ﺍﻟﺒﻴﻮﻟﻮﺟﻴﺎ‪ ،‬ﻣﻦ‬
‫ﺃﺟﻞ ﻗﻴﺎﺱ ﺗﻄﻮﺭ ﳕﻮ ﺑﻌﺾ ﺍﻟﺴﻠﻊ ﺍﻻﺳﺘﻬﻼﻛﻴﺔ ﺍﳌﻌﻤﺮﺓ ﺍﻟﱵ ﺗﺼﻞ ﰲ ﻳﻮﻡ ﻣﺎ ﺇﱃ‬
‫ﻣﺴﺘﻮﻯ ﺍﻹﺷﺒﺎﻉ‪ ،‬ﻭﺫﻟﻚ ﻋﻠﻰ ﻏﺮﺍﺭ ﺍﺳﺘﻌﻤﺎﻻ‪‬ﺎ ﺑﻴﻮﻟﻮﺟﻴﺎ ﻟﻘﻴﺎﺱ ﳕﻮ ﺑﻌﺾ‬
‫ﺍﳊﺸﺮﺍﺕ ﰲ ﺍ‪‬ﺎﻻﺕ ﺍﳌﺨﱪﻳﺔ ﺍﳌﻐﻠﻘﺔ‪.‬‬
‫ﺷﺎﻉ ﺍﺳﺘﻌﻤﺎﻝ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﰲ ﻛﺜﲑ ﻣﻦ ﺍﻟﺪﻭﻝ ﻟﻘﻴﺎﺱ ﺗﻄﻮﺭ ﺑﻌﺾ ﺍﳌﺘﻐﲑﺍﺕ‬
‫ﻭﳏﺎﻭﻟﺔ ﺍﻟﺘﻨﺒﺆ ﺑﺴﻠﻮﻛﻴﺎ‪‬ﺎ ﺍﳌﺴﺘﻘﺒﻠﻴﺔ‪ .‬ﻓﺎﺳﺘﻌﻤﻠﺖ ﰲ ﻛﻞ ﻣﻦ ﺍﻹﲢﺎﺩ ﺍﻟﺴﻮﻓﻴﱵ‬
‫ﺳﺎﺑﻘﺎﹰ ﺃﺛﻨﺎﺀ ﻭﺿﻌﻪ ﺧﻄﺘﻪ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﻷﻭﱃ ﲢﺖ ﺍﻟﻨﻈﺎﻡ ﺍﻻﺷﺘﺮﺍﻛﻲ‪ ،‬ﻛﻤﺎ‬
‫ﺍﺳﺘﻌﻤﻠﺖ ﻟﻘﻴﺎﺱ ﻣﺴﺘﻮﻳﺎﺕ ﺍﻻﺳﺘﻬﻼﻙ ﰲ ﺍﻟﺪﻭﻝ ﺍﻹﺳﻼﻣﻴﺔ ﺍﳌﻌﺎﺻﺮﺓ ‪ ،3‬ﻛﻮﻥ‬
‫ﻫﺬﻩ ﺍﻟﺒﻠﺪﺍﻥ ﻛﺎﻧﺖ ﺗﻌﻴﺶ ﺣﺎﻟﺔ ﺍﺳﺘﻌﻤﺎﺭ ﻭﺍﺳﺘﻨﺰﺍﻑ ﻟﺜﺮﻭﺍ‪‬ﺎ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ‪ ،‬ﻓﻌﻨﺪ‬
‫ﲢﺮﺭﻫﺎ ﻋﺎﺩﺕ ﻋﻠﻴﻬﺎ ﻫﺬﻩ ﺍﻟﺜﺮﻭﺍﺕ ﺍﳌﺴﻠﻮﺑﺔ ﺳﺎﺑﻘﺎ ﲟﻮﺍﺭﺩ ﻣﺎﻟﻴﺔ ﻣﻌﺘﱪﺓ ﺟﻌﻠﺖ ﻣﻦ‬
‫ﺳﻠﻮﻛﻬﺎ ﺍﻻﺳﺘﻬﻼﻛﻲ ﻳﻨﻄﺒﻖ ﻋﻠﻰ ﺷﻜﻞ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ‪.‬‬

‫ﺍﻟﻌﺰﻳﺰ‪،‬ﻡ‪1989 ،1‬‬ ‫‪ -3‬ﳐﺘﺎﺭ ﳏﻤﺪ ﻣﺘﻮﱄ‪ ،‬ﳎﻠﺔ ﺍﻻﻗﺘﺼﺎﺩ ﺍﻹﺳﻼﻣﻲ‪ ،‬ﺟﺎﻣﻌﺔ ﺍﳌﻠﻚ ﻋﺒﺪ‬

‫‪77‬‬
 ‫ﺗﻘﻴ‪‬ﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ‪:‬‬
‫‪Model Validation‬‬
‫ﰲ ﻧﻔﺲ ﺍﻟﺴﻴﺎﻕ ﺍﳌﺘﻌﻠﻖ ﺑﻜﺸﻒ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﳋﺎﺻﺔ ﺑﺎﻻﲡﺎﻩ‬
‫ﺍﻟﻌﺎﻡ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﻏﲑ ﺍﳊﺮﺓ‪ ،‬ﳝﻜﻦ ﺍﻻﺳﺘﻌﺎﻧﺔ ﺑﺄﺳﻠﻮﺏ ﺍﻟﺘﻘﻴﻴﻢ ﳌﻌﺮﻓﺔ‬
‫ﻭﺟﻮﺩ ﺃﻭ ﻋﺪﻡ ﻭﺟﻮﺩ ﻫﺬﻩ ﺍﳌﺮﻛﺒﺔ‪.‬‬
‫ﻗﺒﻞ ﺍﻻﻧﺘﻘﺎﻝ ﺇﱃ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﻧﻠﺠﺄ ﺇﱃ ﺗﻘﻴﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ ﻣﻦ ﺣﻴﺚ ﺟﻮﺩﺓ‬
‫ﺍﻟﺘﻮﻓﻴﻖ ﻭﺍﳌﻌﻨﻮﻳﺔ‪ ،‬ﰒ ﰲ ﺍﻷﺧﲑ ﻗﻴﺎﺱ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ ﺩﺍﺧﻞ ﺍﻟﻌﻴﻨﺔ ﺍﻟﺬﻱ ﺳﻨﻠﺠﺄ ﺇﻟﻴﻪ ﰲ‬
‫‪‬ﺎﻳﺔ ﺍﳌﻄﺎﻑ‪.‬‬
‫ﺑﺎﻓﺘﺮﺍﺽ ﺍﻧﻪ ﻟﺪﻳﻨﺎ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳋﻄﻲ ﺍﻟﻮﺍﺭﺩ ﰲ ﺍﳌﻌﺎﺩﻟﺔ )‪ (3.1‬ﻭﲢﺖ‬
‫ﺍﻟﻔﺮﺿﻴﺎﺕ ﺍﻟﻜﻼﺳﻴﻜﻴﺔ ﳝﻜﻦ ﺇﺟﺮﺍﺀ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫ﺍ‪ -‬ﺟﻮﺩﺓ ﺍﻟﺘﻮﻓﻴﻖ )‪ :(Goodness of fit‬ﻫﻮ ﻣﺮﺑﻊ ﻣﻌﺎﻣﻞ ﺍﻻﺭﺗﺒﺎﻁ ﺍﳋﻄﻲ‬
‫ﺍﳌﺘﻌﺪﺩ‪ ،‬ﻭﻫﻮ ﻳﻌﺮﻑ ﺑﺄﻧﻪ ﻋﺒﺎﺭﺓ ﻋﻦ ﻧﺴﺒﺔ ﻣﺮﺑﻊ ﺍﻻﳓﺮﺍﻓﺎﺕ ﺍﳌﺸﺮﻭﺣﺔ ﺇﱃ ﻣﺮﺑﻊ‬
‫ﺍﻻﳓﺮﺍﻓﺎﺕ ﺍﻟﻜﻠﻴﺔ ﺍﳌﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ ﺍﻟﻌﻤﻠﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪T‬‬

‫‪åe‬‬ ‫‪2‬‬
‫‪t‬‬
‫)‪(3.20‬‬
‫‪t=1‬‬
‫‪R‬‬ ‫‪2‬‬
‫‪= 1-‬‬ ‫‪T‬‬ ‫_‬
‫‪å(y‬‬ ‫‪t‬‬ ‫‪- y )2‬‬
‫‪t =1‬‬

‫ﰲ ﺣﺎﻟﺔ ﺗﻮﻓﺮ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﻋﻠﻰ ﺣﺪ‪ ‬ﺛﺎﺑﺖ‪ (Intercept) ‬ﻓﺈﻥ ﻫﺬﺍ ﺍﳌﻌﻴﺎﺭ‬
‫ﳜﻀﻊ ﻟﻠﺸﺮﻁ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪0£ R2 £1‬‬
‫‪T‬‬
‫)‪(Residual Sum of Squares‬‬ ‫ﺣﻴﺚ ‪ å e 2t‬ﳝﺜﻞ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‬
‫‪t =1‬‬
‫ﻭﺍﻟﱵ ﻧﺮﻣﺰ ﳍﺎ ﺍﺧﺘﺼﺎﺭ‪‬ﺍ ﺑـ )‪.(RSS‬‬
‫ﻳﻜﻮﻥ ﺍﻟﻨﻤﻮﺫﺝ ﻣﻘﺒﻮﻻﹰ ﻛﻠﻤﺎ ﺃﻗﺘﺮﺏ ‪ R 2‬ﻣﻦ ﺍﻟﻮﺍﺣﺪ ﻭﺗﻘﻞ ﺍﻟﺮﻏﺒﺔ ﻓﻴﻪ‬
‫ﻛﻠﻤﺎ ﺍﺑﺘﻌﺪ ﻋﻦ ﻫﺬﺍ ﺍﳌﻘﺪﺍﺭ ﻭﺍﻗﺘﺮﺏ ﻣﻦ ﺍﻟﺼﻔﺮ‪ .‬ﻭﻛﻮﻧﻪ ﻳﺰﺩﺍﺩ ﺑﺰﻳﺎﺩﺓ ﺍﳌﺘﻐﲑﺍﺕ‬

‫‪78‬‬
‫ﺍﻟﺸﺎﺭﺣﺔ ﺇﱃ ﺍﻟﻨﻤﻮﺫﺝ ﺣﱴ ﻭﺇﻥ ﱂ ﻳﻜﻦ ﳍﺎ ﻋﻼﻗﺔ ﺑﺎﻟﻈﺎﻫﺮﺓ ﺍﳌﺪﺭﻭﺳﺔ‪ ،‬ﻓﺈﻥ‬
‫ﺍﻹﺣﺼﺎﺋﻴﲔ ﻳﻔﻀﻠﻮﻥ ﺍﺳﺘﺒﺪﺍﻟﻪ ﲟﻘﻴﺎﺱ ﺟﻮﺩﺓ ﺍﻟﺘﻮﻓﻴﻖ ﺍﳌﻌﺪﻝ ﺍﻟﺘﺎﱄ‪:‬‬
‫_‬
‫)‪( T-1‬‬ ‫)‪(3.21‬‬
‫‪R2 =1-‬‬ ‫) ‪(1- R2‬‬
‫‪T- k‬‬
‫ﺃﻳﻦ ﳝﺜﻞ ‪ k‬ﻋﺪﺩ ﻣﻌﻠﻤﺎﺕ ﺍﳌﻌﺎﺩﻟﺔ‪ .‬ﻭﻫﺬﺍ ﺍﻷﺧﲑ ﻗﺪ ﻳﺰﺩﺍﺩ ﺃﻭ ﻳﻨﻘﺺ‬
‫ﺑﺈﺿﺎﻓﺔ ﻛﻞ ﻣﺘﻐﲑ ﺟﺪﻳﺪ ﻣﺴﺘﻘﻞ ﺇﻟﻴﻬﺎ‪ ،‬ﻭﻋﻤﻠﻴﺎﹰ ﻧﻔﻀﻞ ﺃﻥ ﻳﻜﻮﻥ ﻫﺬﺍ ﺍﻷﺧﲑ‬
‫ﺃﻗﺮﺏ ﻣﺎ ﳝﻜﻦ ﻣﻦ ﺍﻟﻮﺍﺣﺪ )‪.(1‬‬

‫ﻧﺸﲑ ﺇﱃ ﺃﻥ ﺍﻟﻜﺜﲑ ﻣﻦ ﺍﻹﺣﺼﺎﺋﻴﲔ ﻻ ﻳﻌﲑﻭﻥ ﻫﺬﺍ ﺍﳌﻌﺎﻣﻞ ﺍﻫﺘﻤﺎﻣﺎ ﻛﺒﲑﹰﺍ‬

‫ﲝﻜﻢ ﺃﻧﻪ ﻗﺪ ﻳﻜﻮﻥ ﻣﻐﻠﻄﺎﹰ ﰲ ﻛﺜﲑ ﻣﻦ ﺍﳊﺎﻻﺕ ﺧﺎﺻﺔ ﻋﻨﺪ ﺧﻀﻮﻉ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﺍﻟﺰﻣﻨﻴﺔ ﻟﺘﺤﻮﻳﻼﺕ ﺭﻳﺎﺿﻴﺔ ﻣﺜﻞ ﻃﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ‪ .‬ﻛﻤﺎ ﻗﺪ ﻳﺄﺧﺬ ﻗﻴ‪‬ﻤﺎﹰ ﺧﺎﺭﺝ ﳎﺎﻟﻪ‬
‫ﺍﶈﺪﺩ ﻋﻨﺪ ﺗﻐﻴﻴﺐ ﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ ﻣﻦ ﺍﻟﻨﻤﻮﺫﺝ‪.‬‬

‫ﺏ‪ -‬ﺍﺧﺘﺒﺎﺭﺍﺕ ﺍﳌﻌﻨﻮﻳﺔ‪:‬‬

‫ﻻﺧﺘﺒﺎﺭ ﻣﻌﻨﻮﻳﺔ ﺍﳌﻌﻠﻤﺎﺕ ﻛﻞ ﻭﺍﺣﺪﺓ ﻋﻠﻰ ﺣﺪﺍ ﰲ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳋﻄﻲ‬
‫ﺍﻟﺴﺎﺑﻖ‪ ،‬ﻧﻠﺠﺄ ﺇﱃ ﺍﻹﺣﺼﺎﺀﺓ ﺍﻟﺸﻬﲑﺓ )‪ (t- Statistics‬ﻟﻠﻤﻌﻠﻤﺘﲔ ﺍﳌﻌﻄﺎﺓ ﺑﺎﻟﻌﻼﻗﺔ‬
‫ﺍﻟﺘﺎﻟﻴﺔ ﻭﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪:‬‬
‫‪aˆ - a t 1‬‬
‫= ‪ta‬‬ ‫¾‬
‫) ‪¾®( 2 a , T - k‬‬ ‫)‪(3.22‬‬
‫)ˆ‪se(a‬‬
‫‪bˆ - b t 1‬‬ ‫)‪(3.23‬‬
‫= ‪tb‬‬ ‫¾‬
‫) ‪¾® ( 2 a , T - k‬‬
‫)ˆ‪se (b‬‬
‫ﺃﻳﻦ ‪T‬ﻭ ‪ k‬ﲤﺜﻼﻥ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﳌﺴﺘﺨﺪﻣﺔ ﰲ ﺍﻟﺪﺭﺍﺳﺔ ﻭﻋﺪﺩ ﻣﻌﻠﻤﺎﺕ‬
‫^‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪ ،‬ﺑﻴﻨﻤﺎ ) ‪ Se ( b j‬ﲤﺜﻞ ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ ﻟﻠﻤﻌﻠﻤﺔ ﺍﳌﻌﻨﻴﺔ‪،‬‬
‫ﺃﻳﻦ ﻳﺄﺧﺬ ﰲ ﳕﻮﺫﺝ ﺧﻄﻲ ﺑﺴﻴﻂ ﺍﻟﺸﻜﻞ )‪ (3.1‬ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪79‬‬
 ‫^‬
‫‪Se ( a ) = s‬‬ ‫‪2‬‬ ‫‪åt‬‬ ‫‪2‬‬

‫_‬
‫) ‪Tå ( t - t‬‬ ‫‪2‬‬

‫_‬
‫‪1‬‬ ‫‪t‬‬ ‫)‪(3.24‬‬
‫‪=s‬‬ ‫‪+‬‬ ‫_‬
‫‪T‬‬
‫) ‪å( t - t‬‬ ‫‪2‬‬

‫^‬ ‫‪s2‬‬ ‫)‪(3.25‬‬

‫= ) ‪Se (b‬‬ ‫_‬

‫) ‪å (t - t‬‬ ‫‪2‬‬

‫‪s‬‬ ‫‪2‬‬
‫=‬
‫‪åe‬‬ ‫‪2‬‬
‫‪t‬‬
‫=‬
‫‪åe‬‬ ‫‪2‬‬
‫‪t‬‬
‫ﺣﻴﺚ‬
‫‪T -k‬‬ ‫‪T -2‬‬

‫ﺻﻴﻐﺔ ﺍﻻﺧﺘﺒﺎﺭ‪:‬‬
‫ﻧﺴﺘﺤﻀﺮ ﺍﻟﻌﻼﻗﺔ )‪ (3.17‬ﺃﻳﻦ ﻧﺮﻳﺪ ﺍﺧﺘﺒﺎﺭ ﻣﻌﻨﻮﻳﺔ ﺍﳌﻌﻠﻤﺔ ‪ b 1‬ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪H 0 :b 1 = 0‬‬

‫‪H A :b 1 ¹ 0‬‬
‫ﺗﺼﺒﺢ ﺍﻹﺣﺼﺎﺀﺓ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫^‬
‫‪b1‬‬
‫= ‪t1‬‬ ‫^‬
‫) ‪Se ( b 1‬‬
‫ﻭﺗﻜﻮﻥ ﺍﳌﻌﻠﻤﺔ ﺍﳌﻌﻨﻴﺔ ﻣﻌﻨﻮﻳﺔ ﺇﺫﺍ ﻛﺎﻧﺖ ‪:‬‬
‫‪4‬‬

‫‪t1 > t Tab = t 1‬‬

‫‪a ,T - k‬‬
‫‪2‬‬

‫‪. a=5%‬‬ ‫‪ t Tab = t 1‬ﲤﺜﻞ ﺗﻠﻚ ﺍ‪‬ﺪﻭﻟﺔ ﻭﻧﻔﻀﻞ ﺃﻥ ﺗﻜﻮﻥ‬ ‫‪-4‬‬

‫‪a ,T - k‬‬
‫‪2‬‬

‫‪80‬‬
‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‪ ،‬ﺗﺘﺤﻘﻖ ﻣﻌﻨﻮﻳﺔ ﺍﳌﻌﻠﻤﺔ ﳌﺎ ﺗﻜﻮﻥ ‪ . t cal > t Tab‬ﻛﻤﺎ ﻧﺴﺘﻌﲔ‬
‫ﺑﺈﺣﺼﺎﺀﺓ )‪ (Fisher‬ﻋﻨﺪ ﺍﺧﺘﺒﺎﺭ ﻣﻌﻨﻮﻳﺔ ﲨﻠﺔ ﻣﻦ ﺍﳌﻌﻠﻤﺎﺕ ﺃﻧﻴﺎﹰ ﺍﳌﻌﻄﺎﺓ ﻭﻓﻖ‬
‫ﺍﻟﻌﻼﻗﺔ ﺍﻹﺣﺼﺎﺋﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪( RRSS - URSS ) / m F‬‬

‫= ‪Fcal‬‬ ‫)‪¾¾® ( m, T - k ) (3.26‬‬
‫‪URSS / T - k‬‬
‫‪RRSS‬‬ ‫ﺍﻟﱵ ﺗﺘﺒﻊ ﺗﻮﺯﻳﻊ ﻓﻴﺸﺮ ﺏ)‪ ( m‬ﻭ ) ‪ (T - k‬ﺩﺭﺟﺎﺕ ﺣﺮ‪‬ﻳﺔ ﺑﻴﻨﻤﺎ‬
‫ﻭ ‪ URSS‬ﲤﺜﻞ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ﺍﳌﻘﻴﺪﺓ ﲢﺖ ﺍﻟﻔﺮﺿﻴﺔ ‪ H 0‬ﻭﻏﲑ ﺍﳌﻘﻴﺪﺓ‬
‫ﲢﺖ ‪ H A‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ ﺃﻣﺎ )‪ (m‬ﺗﻌﱪ ﻋﻦ ﻋﺪﺩ ﺍﻟﻘﻴﻮﺩ‪.‬‬
‫ﻓﺈﺫﺍ ﻛﺎﻧﺖ ﻫﺬﻩ ﺍﳌﻌﻠﻤﺎﺕ ﺫﺍﺕ ﻣﻌﻨﻮﻳﺔ‪ ،‬ﻧﺘﻮﻗﻊ ﺃﻥ ﺗﻜﻮﻥ ﻗﻴﻤﺔ ﺍﻹﺣﺼﺎﺀﺓ‬
‫ﺍﶈﺴﻮﺑﺔ ﺃﻛﱪ ﻣﻦ ﺍ‪‬ﺪﻭﻟﺔ ﺃﻱ‪:‬‬

‫) ‪Fcal > F( m, T - k‬‬

‫ﻟﺘﻮﺿﻴﺢ ﺍﻟﻔﻜﺮﺓ ﻧﻌﻮﺩ ﺇﱃ ﻣﻌﻄﻴﺎﺕ ﺍﳌﺜﺎﻝ )‪ (12‬ﺟﺪﻭﻝ )‪ (17‬ﻭﺍﳌﻤﺜﻠﺔ ﺑﻴﺎﻧﻴﹰﺎ‬

‫ﰲ ﺍﻟﺸﻜﻞ )‪ (17‬ﻭﺍﻟﺬﻱ ﻧﺴﺘﻨﺘﺞ ﻣﻨﻪ ﺇﻣﻜﺎﻧﻴﺔ ﺍﻟﺘﻌﺒﲑ ﻋﻦ ﺫﻟﻚ ﺍﻟﺸﻜﻞ ﺃﻳﻀﺎﹰ ﻭﻓﻖ‬
‫ﺍﳌﻌﺎﺩﻟﺔ )‪ (3.16‬ﻭﺍﻟﱵ ﻧﻌﻴﺪ ﻛﺘﺎﺑﺘﻬﺎ ﻟﻠﻤﺮﺓ ﺍﻟﺜﺎﻧﻴﺔ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪y t = b0 + b1 t + b2 t 2 + u t‬‬

‫ﻋﻨﺪ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ‪ ،‬ﰎ ﺍﻟﺘﻮﺻﻞ‬
‫ﺇﱃ ﺍﻟﻨﺘﺎﺋﺞ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫^‬
‫‪y t = 391. 86 - 103. 23 t + 21. 13 t 2‬‬
‫) ‪( 52. 73) ( 30. 22‬‬ ‫)‪( 3. 69‬‬

‫‪R 2 = 0. 97‬‬ ‫‪åe‬‬ ‫‪t‬‬

‫‪2‬‬
‫‪= 4579 .38‬‬

‫‪81‬‬
 ‫ﻣﺎ ﺑﲔ ﺍﻟﻘﻮﺳﲔ ﲤﺜﻞ ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ ﻟﻠﻤﻌﻠﻤﺔ ‪.‬‬
‫ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﺍﻟﻮﻗﻮﻑ ﻋﻠﻰ ﻣﻌﻨﻮﻳﺔ ﺍﳌﻌﻠﻤﺔ ‪ b 2‬ﻣﺜﻼﹰ ﺑﺎﺳﺘﺨﺪﺍﻡ ﻛﻞ ﻣﻦ‬
‫ﺍﺧﺘﺒﺎﺭﻱ ﺳﺘﻴﻮﺩﻧﺖ ﻭﻓﻴﺸﺮ )‪ (t-Student &Fisher statistics‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‬
‫ﻛﻤﺎ ﻳﻠﻲ ‪:‬‬
‫ﻧﺴﺘﻬﻞ ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ ﺑﺘﻄﺒﻴﻖ ﺍﺧﺘﺒﺎﺭ ﺳﺘﻴﻮﺩﻧﺖ ﻋﻠﻰ ﺍﳌﻌﻠﻤﺔ ‪ b2‬ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪H 0 :b2 = 0‬‬

‫‪H a :b2 ¹ 0‬‬
‫^‬

‫= ‪t2‬‬
‫‪b2‬‬
‫^‬
‫=‬
‫‪21. 13 - 0‬‬
‫‪= 5. 72‬‬ ‫ﻣﻨﻪ‬
‫) ‪Se ( b 2‬‬ ‫‪3. 69‬‬

‫‪t ( 5%, 4 ) = t Tab = 2. 132‬‬ ‫ﻭ‬

‫ﲟﻘﺎﺭﻧﺔ ﺍﻹﺣﺼﺎﺀﺓ ﺍﶈﺴﻮﺑﺔ ﺃﻋﻼﻩ ﻣﻊ ﺍ‪‬ﺪﻭﻟﺔ ﻳﺘﻢ ﺭﻓﺾ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ‪،‬‬
‫ﲟﻌﲎ ﺃﻥ ﺍﳌﺘﻐﲑ ﺫﻭ ﺍﳌﻌﻠﻤﺔ ‪ b 2‬ﻟﻪ ﺩﻻﻟﺔ ﺇﺣﺼﺎﺋﻴﺔ‪.‬‬
‫ﺑﻌﺪ ﺍﺧﺘﺒﺎﺭ ﻣﻌﻨﻮﻳﺔ ﻛﻞ ﻣﻦ ‪ b 0‬ﻭ ‪ b 1‬ﺃﻳﻦ ﰎﹶ ﺭﻓﺾ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ﰲ ﻛﻼ‬
‫ﺍﳊﺎﻟﺘﲔ‪ ،‬ﻧﺴﺘﻄﻴﻊ ﺍﻟﻘﻮﻝ ﺑﺄﻧﻪ ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻦ ﻣﻌﻄﻴﺎﺕ ﺍﳉﺪﻭﻝ )‪ (17‬ﺑﺪﺍﻟﺔ ﺍﻟﻘﻄﻊ‬
‫ﺍﳌﻜﺎﻓﺊ ﺍﳌﺒﻴﻨﺔ ﰲ ﺍﻟﺸﻜﻞ )‪.(17‬‬
‫ﻳﺘﻤﻴﺰ ﺍﻻﺧﺘﺒﺎﺭ ﺍﻟﺴﺎﺑﻖ ﺑﺴﻬﻮﻟﺔ ﺍﻟﺘﻄﺒﻴﻖ‪ ،‬ﺣﻴﺚ ﻳﺘﻄﻠﺐ ﺗﻘﺪﻳﺮ ﺍﻟﻨﻤﻮﺫﺝ ﻣﺮﺓ‬
‫ﻭﺍﺣﺪﺓ ﻓﻘﻂ ﻭﲢﺖ ﺍﻟﻔﺮﺿﻴﺔ ﺍﻟﺒﺪﻳﻠﺔ ‪ ، H a‬ﺑﻴﻨﻤﺎ ﺍﺧﺘﺒﺎﺭ ﻓﻴﺸﺮ ﻟﻠﻤﻌﻨﻮﻳﺔ ﻳﺘﻄﻠﺐ‬
‫ﺗﻘﺪﻳﺮ ﺍﻟﻨﻤﻮﺫﺝ ﻣﺮﺗﲔ‪ ،‬ﻭﺍﺣﺪﺓ ﲢﺖ ﺍﻟﻔﺮﺿﻴﺔ ‪ H a‬ﻫﻮ ﻧﻔﺲ ﺍﻟﺘﻘﺪﻳﺮ ﺃﻋﻼﻩ‪ ،‬ﺃﻳﻦ‬
‫ﻧﺄﺧﺬ ﻣﻨﻪ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ﻓﻘﻂ ﻭﺍﻟﱵ ﻧﺴﻤﻴﻬﺎ ﺑـ ‪ ، URSS‬ﻭﺍﳌﺮﺓ‬
‫ﺍﻷﺧﺮﻯ ﲢﺖ ‪ ، H 0‬ﺣﻴﺚ ‪ b 2‬ﻣﻌﺪﻭﻣﺔ ﺃﻱ‪:‬‬

‫‪y t = b0 + b1 t + u t‬‬

‫‪82‬‬
 ‫ﻋﻨﺪ ﺗﻘﺪﻳﺮﻩ ﻫﺬﻩ ﺍﳌﺮﺓ ﳓﺼﻞ ﻋﻠﻰ ﺍﻟﻨﺘﺎﺋﺞ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫^‬
‫‪y t = 138. 28 + 65. 82 t‬‬
‫) ‪( 77. 53 ) ( 17. 33‬‬

‫‪åe‬‬ ‫‪2‬‬
‫‪t‬‬ ‫‪= 42086. 82‬‬

‫ﻧﺴﻤﻲ ﳎﻤﻮﻉ ﻫﺬﻩ ﺍﳌﺮﺑﻌﺎﺕ ﺑـ ‪ ، RRSS‬ﻭﻣﻨﻪ ﻳﺘﻢ ﺗﺸﻜﻴﻞ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ‬

‫ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪(42086.82 - 4579.38) / 1‬‬
‫= ‪Fcal‬‬ ‫‪= 32.76‬‬
‫‪4579.38 / 4‬‬

‫ﻭﲟﻘﺎﺭﻧﺔ ﻫﺬﺍ ﺍﻷﺧﲑ ﻣﻊ ﺗﻠﻚ ﺍ‪‬ﺪﻭﻟﺔ ‪ ، f Tab = 7. 71‬ﻧﻼﺣﻆ ﺃﻥ ﺍﶈﺴﻮﺑﺔ‬

‫ﺃﻛﱪ ﻣﻦ ﺍ‪‬ﺪﻭﻟﺔ ﻭﺑﺎﻟﺘﺎﱄ ﺭﻓﺾ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ‪ ،‬ﻭﻫﻮ ﻣﺎ ﻳﺘﻮﺍﻓﻖ ﻣﻊ ﻗﺮﺍﺭ ﺍﺧﺘﺒﺎﺭ‬
‫ﺳﺘﻴﻮﺩﻧﺖ ﻷﻥ ‪. Fcal = t 2cal 5‬‬
‫ﻧﺬﻛﺮ ﺑﺄﻧﻪ ﻳﻔﻀﻞ ﺍﺳﺘﺨﺪﺍﻡ ﺍﺧﺘﺒﺎﺭ ﻓﻴﺸﺮ ﻻﺧﺘﺒﺎﺭ ﺃﻛﺜﺮ ﻣﻦ ﻣﻌﻠﻤﺔ ﻭﺍﺣﺪﺓ‪،‬‬
‫ﻟﻘﺪ ﻛﺎﻥ ﺍﺳﺘﺨﺪﺍﻣﻪ ﻫﻨﺎ ﺑﻐﺮﺽ ﺗﻮﺿﻴﺢ ﻛﻴﻔﻴﺔ ﺍﻻﺳﺘﻌﻤﺎﻝ‪.‬‬
‫ﺇﻥ ﺍﳌﻼﺣﻈﺔ ﺍﳌﻴﺪﺍﻧﻴﺔ ﺗﺒﲔ ﺃﻥ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﳌﺪﺭﻭﺳﺔ ﻻ ﺗﺄﺧﺬ ﰲ‬
‫ﺍﻟﻐﺎﻟﺐ ﺍﻟﺸﻜﻞ ﺍﻟﻨﻈﺎﻣﻲ ﺑﻘﺪﺭ ﻣﺎ ﻳﻜﻮﻥ ﺍﻟﻄﺎﺑﻊ ﺍﻟﻌﺸﻮﺍﺋﻲ ﺑﺎﺭﺯ‪‬ﺍ ﻓﻴﻬﺎ‪ ،‬ﻭﳍﺬﺍ‬
‫ﻓﺪﺭﺍﺳﺔ ﻫﺬﻩ ﺍﻟﺴﻼﺳﻞ ﻭﻓﻖ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺘﻨﺎﻭﻟﺔ ﻗﺪ ﻳﻌﻄﻲ ﻧﺘﺎﺋﺞ ﻏﲑ ﻣﺮﻏﻮﺏ ﻓﻴﻬﺎ‪،‬‬
‫ﺑﺴﺒﺐ ﺃﻥ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ‪‬ﻤﻞ ﺍﻟﺬﺑﺬﺑﺎﺕ ﺍﳊﺎﺩﺓ ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﻭﺍﻫﺘﻤﺎﻣﻬﺎ ﺑﺎﳌﺮﻛﺒﺔ ﺍﻟﻨﻈﺎﻣﻴﺔ ﻓﻘﻂ ﺟﻌﻠﻬﺎ ﺗﻔﻘﺪ ﳎﻤﻮﻋﺔ ﻣﻌﺘﱪﺓ ﻣﻦ ﺍﳌﻌﻠﻮﻣﺎﺕ ﻗﺪ‬
‫ﺗﻔﻴﺪ ﺑﺸﻜﻞ ﻛﺒﲑ ﰲ ﲢﺴﲔ ﺍﻟﺘﻨﺒﺆ‪ ،‬ﻭﻋﺪﻡ ﺗﻌﺮﻳﺾ ﺍﳌﺆﺳﺴﺔ ﻣﺜﻼ ﻻ ﻟﻔﺮﺻﺔ ﺿﺎﺋﻌﺔ‬
‫)ﰲ ﺣﺎﻟﺔ ﻣﺎ ﺇﺫﺍ ﻛﺎﻥ ﺇﻧﺘﺎﺟﻬﺎ ﺃﻗﻞ ﻣﻦ ﺍﳌﺴﺘﻮﻯ ﺍﳌﻄﻠﻮﺏ(‪ ،‬ﻭﻻ ﺇﱃ ﲣﺰﻳﻦ ﻧﺴﺒﺔ‬

‫‪ -5‬ﻳﺴﺘﻄﻴﻊ ﺍﻟﻘﺎﺭﺉ ﺍﻟﺘﺄﻛﺪ ﻣﻦ ﺫﻟﻚ ﺭﻗﻤﻴﺎﹰ‪.‬‬

‫‪83‬‬
‫ﻋﺎﻟﻴﺔ ﻣﻦ ﺇﻧﺘﺎﺟﻬﺎ )ﺍﻹﻧﺘﺎﺝ ﺃﻛﱪ ﻣﻦ ﺍﻟﻄﻠﺐ(‪ ،‬ﳑﺎ ﻳﺘﺴﺒﺐ ﰲ ﺗﻜﺎﻟﻴﻒ ﺇﺿﺎﻓﻴﺔ ﻣﻦ‬
‫ﺟﺮﺍﺀ ﻋﻤﻠﻴﺔ ﺍﻟﺘﺨﺰﻳﻦ ﻭﺍﶈﺎﻓﻈﺔ ﻋﻠﻰ ﺍﻟﻨﺎﺗﺞ ﰲ ﺣﺎﻟﺔ ﺟﻴﺪﺓ ﰲ ﺃﺣﺴﻦ ﺍﻷﺣﻮﺍﻝ‪.‬‬

‫‪Adaptive Models‬‬ ‫‪ -2‬ﺍﻟﺘﻨﺒﺆ ﻭﺍﻟﺘﻤﻬﻴﺪ ﺑﺎﻟﻨﻤﺎﺫﺝ ﺍﳌﻜﻴ‪‬ﻔﺔ‬

‫ﻧﻘﺼﺪ ﺑﺘﻘﻨﻴﺔ ﺍﻟﺘﻤﻬﻴﺪ ﻋﻤﻠﻴﺔ ﺍﻹﺯﺍﻟﺔ ﺃﻭ ﺍﻟﺘﺨﻔﻴﻒ ﻣﻦ ﺣﺪﺓ ﺍﻟﺬﺑﺬﺑﺎﺕ‬
‫ﺍﻟﻘﺼﲑﺓ ﺍﳌﺪﻯ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ .‬ﺗﺴﻤﺢ ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ ﺑﺘﺒﺴﻴﻂ ﺷﻜﻞ ﻣﻨﺤﲎ‬
‫ﺍﻟﻈﺎﻫﺮﺓ ﺍﳌﺪﺭﻭﺳﺔ ﻭﻣﻨﻪ ﺗﺴﻬﻴﻞ ﻋﻤﻠﻴﺔ ﺣﺼﺮ ﺍﳌﺮﻛﺒﺎﺕ ﺍﻟﻨﻈﺎﻣﻴﺔ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﻭﻣﻨﻪ‬
‫ﺗﺴﻬﻴﻞ ﺩﺭﺍﺳﺘﻬﺎ‪ .‬ﺑﻴﻨﻤﺎ ﻳﻘﺼﺪ ﺑﺎﻟﺘﻨﺒﺆ ﳏﺎﻭﻟﺔ ﺍﺳﺘﺸﺮﺍﻑ ﺍﳌﺴﺘﻘﺒﻞ ﺍﻟﻘﺮﻳﺐ ﺑﺘﻘﻨﻴﺎﺕ‬
‫ﺑﺴﻴﻄﺔ‪.‬‬
‫ﺍ‪ -‬ﺍﻟﺘﻨﺒﺆ‬
‫ﲤﻴﺰ ﺗﻄﺒﻴﻖ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺴﺎﺑﻘﺔ‪ ،‬ﻋﻠﻰ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﱵ ﺗﱪﺯ ﻓﻴﻬﺎ ﺑﺸﻜﻞ‬
‫ﻭﺍﺿﺢ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﺑﻴﻨﻤﺎ ﻳﺘﻼﺀﻡ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﻣﻊ ﺍﻟﺴﻼﺳﻞ‬
‫ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﱵ ﺗﺘﻤﻴﺰ ﺑﺎﻻﺳﺘﻘﺮﺍﺭ ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﻓﻬﻲ ﺧﺎﻟﻴﺔ ﻣﻦ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ‬
‫ﺍﻟﻌﺎﻡ ﻭﺍﻟﻔﺼﻠﻴﺔ ﰲ ﻣﺮﺣﻠﺔ ﺃﻭﻟﻴﺔ‪ ،‬ﻭﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻳﺘﻢ ﺑﻠﻮﻏﻬﺎ ﺑﻨﻔﺲ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ‬
‫ﺍﻟﺴﺎﺑﻘﺔ ﺍﻟﺬﻛﺮ ﻋﻨﺪ ﳏﺎﻭﻟﺔ ﺍﻟﺘﺄﻛﺪ ﻣﻦ ﻭﺟﻮﺩ ﺃﻭ ﻏﻴﺎﺏ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﻛﺬﺍ‬
‫ﺍﻟﻔﺼﻠﻴﺔ‪ .‬ﺗﺘﻄﻮﺭ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﰲ ﻣﺮﺣﻠﺔ ﻻﺣﻘﺔ ﻟﺘﺘﻼﺀﻡ ﻭﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﲟﺮﻛﺒﺎ‪‬ﺎ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪ ،‬ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﺿﻤﻦ ﳕﺎﺫﺝ ﺁﻧﻴﺔ ﺇﻥ ﺻﺢ‪ ‬ﺍﻟﺘﻌﺒﲑ‪.‬‬
‫‪6‬‬

‫ﻹﻋﻄﺎﺀ ﻓﻜﺮﺓ ﺣﻮﻝ ﺍﳌﻮﺿﻮﻉ‪ ،‬ﻧﺸﲑ ﺇﱃ ﺃﻥ ﳕﻮﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﻟﺴﺎﺑﻖ‬

‫ﺍﻟﺬﻛﺮ ﻳﺄﺧﺬ ﺍﻟﺸﻜﻞ ﺍﻟﻌﺸﻮﺍﺋﻲ ﺇﺫﺍ ﻛﺎﻧﺖ ﻣﻌﻠﻤﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻣﻌﺪﻭﻣﺔ‪ ،‬ﺃﻱ‪:‬‬
‫‪yt = a + e t‬‬
‫‪aˆ = y‬‬ ‫ﻣﻨﻪ‬
‫ˆ‪yˆT+1 = a‬‬ ‫ﺑﺎﻟﺘﺎﱄ‪ ،‬ﻳﻜﻮﻥ ﺍﻟﺘﻨﺒﺆ‪:‬‬

‫‪ - 6‬ﳕﻮﺫﺝ ‪ Holt - Winters‬ﻣﺜﻼﹰ‪.‬‬

‫‪84‬‬
‫ﻫﻮ ﻣﻘﺪﺍﺭ ﺛﺎﺑﺖ ﻋﻨﺪ ﺫﻟﻚ ﺍﳌﺴﺘﻮﻯ )‪ (intercept‬ﰲ ﻛﻞ ﺍﻟﻔﺘﺮﺍﺕ ﺍﻟﺴﺎﺑﻘﺔ‬
‫ﺍﻧﻄﻼﻗﺎﹰ ﻣﻦ ﺍﻟﻔﺘﺮﺓ ﺍﳊﺎﻟﻴﺔ‪ .‬ﻛﻤﺎ ﺃﻥ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﻳﻜﻴﻒ ﻧﻔﺴﻪ ﺃﻟﻴﺎﹰ ﻟﻜﻞ‬
‫ﻭﺿﻊ ﺟﺪﻳﺪ‪ ،‬ﺫﻟﻚ ﻋﻨﺪ ﺗﻮﻓﺮ ﻣﻌﻠﻮﻣﺎﺕ ﺇﺿﺎﻓﻴﺔ ﺟﺪﻳﺪﺓ ﺗﺪﺧﻞ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﺣﻴﺚ ﻳﺘﻢ ﺇﻋﺎﺩﺓ ﺗﻘﻴﻴ‪‬ﻢ ﻣﻌﻠﻤﺎﺕ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﻛﺬﺍ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ‪.‬‬
‫ﻣﻦ ﺑﲔ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﻣﺎ ﻳﻠﻲ‪:‬‬

‫‪ -1‬ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‪:‬‬

‫ﺗﻌﺘﻤﺪ ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﻣﻦ ﺍﻟﻄﺮﻕ ﰲ ﺣﺴﺎﺑﺎ‪‬ﺎ ﻋﻠﻰ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﺳﻮﺍﺀ‬
‫ﻣﻨﻪ ﺍﻟﺒﺴﻴﻂ ﺃﻭ ﺍﳌﺮﺟﺢ‪ ،‬ﻭ ﳝﻜﻦ ﺗﻘﺴﻴﻤﻬﺎ ﺇﱃ‪:‬‬
‫‪ -1-1‬ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﻟﺒﺴﻴﻄﺔ‪:‬‬
‫ﻭﺗﻜﺘﺐ ﻋﻼﻗﺘﻬﺎ ﺍﻟﺮﻳﺎﺿﻴﺔ ﻭﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ ‪:‬‬
‫^‬ ‫‪1‬‬
‫= ‪y T+1‬‬ ‫) ‪( y T + y T - 1 + y T - 2 +... + y T - n + 1‬‬
‫‪n‬‬
‫‪n -1‬‬
‫^‬ ‫‪1‬‬
‫= ‪y T+ 1‬‬
‫‪n‬‬
‫)‪å y ..... (3.27‬‬
‫‪r= 0‬‬
‫‪T- r‬‬

‫ﻓﺈﻥ ﺍﻟﻌﻼﻗﺔ ﺃﻋﻼﻩ ﺗﺼﺒﺢ‪:‬‬ ‫‪n =4‬‬ ‫ﻓﺈﺫﺍ ﻛﺎﻧﺖ‬

‫^‬ ‫) ‪( y T + y T- 1 + y T- 2 + y T- 3‬‬
‫= ‪y T+ 1‬‬
‫‪4‬‬
‫ﺃﻳﻦ‬
‫^‬ ‫) ‪( y T- 1 + y T- 2 + y T- 3 + y T- 4‬‬
‫= ‪yT‬‬
‫‪4‬‬
‫ﻫﺬﺍ ﳝﺜﻞ ﺭﻗﻤﻴﺎ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫^‬ ‫) ‪( y 79 + y 78 + y 77 + y 76‬‬
‫= ‪y 80‬‬
‫‪4‬‬

‫)‪(The most available observation‬‬ ‫ﺃﺣﺪﺙ ﻣﺸﺎﻫﺪﺓ‬ ‫ﺣﻴﺚ ﺗﻌﺘﱪ‬

‫‪y 79‬‬
‫ﳝﻜﻦ ﺍﺳﺘﻌﻤﺎﳍﺎ ﺍﻵﻥ‪.‬‬

‫‪85‬‬
‫ﺇﺿﺎﻓﺔ ﺇﱃ ﻏﻴﺎﺏ ﻃﺮﻳﻘﺔ ﻟﻠﺘﻘﺪﻳﺮ‪ ،‬ﻭﻣﻨﻪ ﻏﻴﺎﺏ ﻭﺳﺎﺋﻞ ﺇﺣﺼﺎﺋﻴﺔ ﻣﻼﺋﻤﺔ‬
‫ﻟﺘﻘﻴﻴﻢ ﻗﻮﺓ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻹﺣﺼﺎﺋﻴﺔ ﻭﺍﻟﺘﻨﺒﺆﻳﺔ‪ ،‬ﻓﺈﻥ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﻳﻨﺘﻘﺪ ﻋﻠﻰ ﺃﺳﺎﺱ ﺃﻧﻪ‬
‫ﻳﻌﺘﱪ ﺃﻥ ﻣﺴﺘﻘﺒﻞ ‪ y‬ﻣﺎ ﻫﻮ ﺇﻻ ﺍﳌﺘﻮﺳﻂ ﺍﳊـﺴﺎﰊ ﻟﻘﻴﻤﻪ ﺍﻟﺴﺎﺑﻘﺔ ﻓﻘﻂ‪ ،‬ﻭﳍﺬﺍ‬
‫ﻛﺎﻥ ﻣﻦ ﺍﻟﻮﺍﺟﺐ ﺗﻌﺪﻳﻠﻪ ﺃﻭ ﺗﻜﻴ‪‬ﻴﻔﻪ ﻟﻴﻌﻄﻲ ﳕﻮﺫﺟ‪‬ﺎ ﺃﻓﻀﻞ ﻳﺄﺧﺬ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ‬
‫ﻫﺬﺍ ﺍﻻﻧﺘﻘﺎﺩ ﻭﻳﻌﻄﻲ ﻭﺯﻥ ﺃﻛﱪ ﻟﺘﺄﺛﲑ ﻗﻴ‪‬ﻢ ‪ y‬ﺍﳊﺎﻟﻴﺔ ﻋﻦ ﺳﺎﺑﻘﺘﻬﺎ‪ ،‬ﻭﻫﻜﺬﺍ ﺗﺮﺍﺟﻌﻴﺎ‬
‫ﺇﱃ ﻧﻘﻄﺔ ﺑﺪﺍﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﺃﻱ ‪ ، y 0 ... ، y t - 2 ، y t -1 ، y t‬ﺇﺿﺎﻓﺔ ﺇﱃ ﺃﻥ‬
‫ﺍﻟﻨﺘﻴﺠﺔ ﺗﺘﻮﻗﻒ ﺑﺎﻷﺳﺎﺱ ﻋﻠﻰ ﺍﻟﻘﻴﻤﺔ ‪ n‬ﺍﻟﱵ ﲣﺘﺎﺭ ﻋﻔﻮﻳﺎﹰ‪ .‬ﳍﺬﺍ ﻳﻘﺘﺮﺡ ﺃﺧﺬﻫﺎ ﰲ‬
‫ﺣﺎﻟﺔ ﺍﺳﺘﻌﻤﺎﻝ ﻣﻌﻄﻴﺎﺕ ﻓﺼﻠﻴﺔ ﻣﺴﺎﻭﻳﺔ ﻷﺭﺑﻌﺔ ) ‪ ( n = 4‬ﺑﻴﻨﻤﺎ ﰲ ﺍﳌﻌﻄﻴﺎﺕ‬
‫ﺍﻟﺸﻬﺮﻳﺔ ﺗﺴﺎﻭﻱ )‪ (12‬ﻭﰲ ﺍﳊﺎﻻﺕ ﺍﻷﺧﺮﻯ ﺗ‪‬ﺨﺘﺎﺭ ﺑﻄﺮﻳﻘﺔ ﻋﻔﻮﻳﺔ ﺃﻳﻀﺎﹰ‪.‬‬
‫‪ -2-1‬ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﳌﺮﺟﺤﺔ ﺍﻷﺳﻴﺔ ‪:‬‬
‫‪7‬‬

‫ﻋﻜﺲ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺴﺎﺑﻖ‪ ،‬ﺍﻟﺬﻱ ﻳﻌﻄﻲ ﻧﻔﺲ ﺍﻟﻮﺯﻥ ﻟﻠﻤﺘﻐﲑ ‪ y‬ﰲ ﳐﺘﻠﻒ‬
‫ﺍﻟﻔﺘﺮﺍﺕ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻓﺈﻥ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﳜﺼﺺ ﺍﻟﻮﺯﻥ ﺍﻷﻛﱪ ﻟﻠﻘﻴ‪‬ﻢ ﺍﳊﺎﻟﻴﺔ ﻋﻦ‬
‫ﺳﺎﺑﻘﺘﻬﺎ ﺑﺸﻜﻞ ﻣﺘﻨﺎﻗﺺ ﺍﻟﺘﺄﺛﲑ‪ .‬ﻧﻌﲏ ‪‬ﺬﻩ ﺍﻟﻔﻜﺮﺓ ﺃﻥ ﻳﻜﻮﻥ ﺗﺄﺛﲑ ‪ y t -1‬ﺃﻛﱪ ﻣﻦ‬
‫ﺗﺄﺛﲑ ‪ y t - S‬ﺣﻴﺚ ‪ ، s > 1‬ﻛﻤﺎ ﺃﻧﻪ ﲣﻠﺺ ﻣﻦ ﺇﺷﻜﺎﻟﻴﺔ ) ‪.( n‬‬
‫ﻭﻳﻜﺘﺐ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺭﻳﺎﺿﻴﺎ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ ‪:‬‬
‫‪¥‬‬
‫‪yˆT +1 = a å (1 - a )r yT - r‬‬ ‫)‪(3.28‬‬
‫‪r =0‬‬

‫ﺣﻴﺚ ‪ ، 0 < a < 1‬ﻭﻫﻲ ﻣﻌﻠﻤﺔ ﺍﻟﺘﻜﻴ‪‬ﻴ‪‬ﻒ‪ ،‬ﻓﻜﻠﻤﺎ ﺍﻗﺘﺮﺑﺖ ﻣﻦ ﺍﻟﻮﺍﺣﺪ‬

‫ﻛﺎﻥ ﺍﻟﺘﻌﺪﻳﻞ ﺃﺳﺮﻉ‪ ،‬ﻭﻛﻠﻤﺎ ﺍﺑﺘﻌﺪﺕ ﻋﻨﻪ ﻛﺎﻥ ﺍﻟﺘﻌﺪﻳﻞ ﺃﺑﻄﺄ‪.‬‬
‫ﻳﻜﻮﻥ ﺍﺳﺘﻌﻤﺎﻝ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﰲ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﻛﺎﻷﰐ‪:‬‬
‫‪2‬‬ ‫‪3‬‬
‫‪yˆT +1 = a yT + a (1 - a ) yT -1 + a (1 - a ) yT - 2 + a (1 - a ) yT -3 + ...‬‬

‫‪¥‬‬
‫‪yˆ T +1 = a å (1 - a ) r yT - r‬‬ ‫)‪(3.29‬‬
‫‪T‬‬ ‫‪r =0‬‬

‫‪7 - Exponentially Weighted moving averages‬‬

‫‪86‬‬
 ‫ﺍﻧﻄﻼﻗﺎ ﻣﻦ ﺍﻟﻔﺘﺮﺓ ﺍﳊﺎﻟﻴﺔ ‪. T‬‬ ‫‪T+1‬‬ ‫ﻟﻠﻔﺘﺮﺓ‬ ‫‪y‬‬ ‫ﻫﻮ ﺗﻨﺒﺆ‬ ‫‪yˆT +1‬‬ ‫ﺃﻳﻦ‬
‫‪T‬‬
‫‪2‬‬ ‫‪3‬‬
‫‪yˆ T +2 = a yˆ T +1 + a (1 - a ) yT + a (1 - a ) yT -1 + a (1 - a ) yT -2 + ...‬‬

‫{‬
‫‪= a yˆ T +1 + (1 - a ) a y + a (1 - a ) y‬‬
‫‪T‬‬ ‫‪T -1‬‬
‫‪+ a (1 - a )2 y‬‬
‫‪T -2‬‬
‫‪+ ...‬‬ ‫}‬
‫^‬ ‫^‬
‫‪= a y T +1 + (1 - a ) y T + 1‬‬ ‫)‪(3.30‬‬

‫ﻣﻦ ﻫﻨﺎ ﻧﺴﺘﻨﺘﺞ ﻭﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ﺃﻥ‪:‬‬

‫‪yˆT +2 = yˆT +1‬‬

‫‪yˆT +L = yˆT +1‬‬ ‫ﻭ‬

‫ﻛﻤﺎ ﻧﻼﺣﻆ‪ ،‬ﻭﺇﺿﺎﻓﺔ ﺇﱃ ﻛﻮﻥ ﺗﻨﺒﺆﺍﺕ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﻏﲑ ﳐﺘﻠﻔﺔ ﻋﻦ‬
‫ﺑﻌﻀﻬﺎ ﺍﻟﺒﻌﺾ‪ ،‬ﻓﺈﻥ ﺍﺳﺘﻌﻤﺎﻝ ﺍﻟﻌﻼﻗﺔ )‪ (3.29‬ﻳﺴﺒﺐ ﻋﺮﺍﻗﻴﻞ ﻛﺜﲑﺓ ﺃﳘﻬﺎ ﻻ‪‬ﺎﺋﻴﺔ‬
‫ﺍﻟﻌﻼﻗﺔ ﺍﻟﺮﻳﺎﺿﻴﺔ‪ ،‬ﻟﺬﻟﻚ ﻭﺟﺐ ﺗﺒﺴﻴﻄﻬﺎ ﻛﺎﻷﰐ‪:‬‬
‫‪¥‬‬
‫‪yˆ T +1 = a å (1 - a ) r yT -r‬‬
‫‪r =0‬‬

‫‪yˆ T +1 = ayT + a (1 - a ) yT -1 + a (1 - a ) 2 yT -2 + a (1 - a ) 3 yT -3 + ... + a (1 - a ) m yT -m ...‬‬

‫‪2‬‬
‫}‪= a yT + (1 - a){a yT -1 + a(1 - a) yT -2 + a(1 - a) yT -3 ...‬‬

‫‪ yˆT‬ﺇﺫﺍ‪:‬‬ ‫‪/ T -1‬‬

‫ﻭﻛﻮﻥ ﺍﳌﻘﺪﺍﺭ ﺍﳌﻮﺟﻮﺩ ﺑﲔ ﺍﻟﻘﻮﺳﲔ ﺍﻟﻜﺒﲑﻳﻦ ﻳﺴﺎﻭﻱ‬
‫‪$y T+ 1 = ay T + ( 1 - a ) $y T/T - 1‬‬ ‫)‪(3.31‬‬

‫‪yˆ T +1 = ayT + (1 - a ) yˆ T‬‬

‫) ‪= yˆ T + a ( yT - yˆ T‬‬

‫) ‪= yˆT - a (et‬‬

‫‪87‬‬
‫ﺗﻌﺘﱪ ﻫﺬﻩ ﺍﻟﻌﻼﻗﺔ ﺍﻷﺧﲑﺓ ﺃﻥ ﺗﻨﺒﺆ ﺍﻟﻔﺘﺮﺓ ﺍﻟﻘﺎﺩﻣﺔ ﻣﺎ ﻫﻮ ﺇﻻ ﺗﻨﺒﺆ ﺍﻟﻔﺘﺮﺓ‬
‫ﺍﻟﺴﺎﺑﻘﺔ ﻣﻀﺎﻓﺎﹰ ﺇﻟﻴﻪ ﻣﻘﺪﺍﺭ ﺍﻟﺘﻌﺪﻳﻞ‪ ،‬ﻭﻫﻮ ﻣﻜﻮﻥ ﻣﻦ ﻧﺴﺒﺔ ﻣﻦ ﺍﻟﺒﻮﺍﻗﻲ‬
‫)‪.(Residuals‬‬

‫ﳝﻜﻦ ﺍﻹﺷﺎﺭﺓ ﻣﻦ ﺧﻼﻝ ﺍﻟﻨﻈﺮ ﺇﱃ ﺍﻟﻌﻼﻗﺔ )‪ (3.27‬ﻭ)‪ (3.31‬ﺃﻧﻪ ﳝﻜﻦ‬

‫ﺍﳊﺼﻮﻝ ﻋﻠﻰ ‪ yˆT +1‬ﺑﻜﻞ ﺳﻬﻮﻟﺔ ﻣﻦ ﺍﻷﻭﱃ‪ ،‬ﻭﺑﻨﻮﻉ ﻣﻦ ﺍﻟﺼﻌﻮﺑﺔ ﻣﻦ ﺍﻟﺜﺎﻧﻴﺔ‪،‬‬
‫ﻧﺘﻴﺠﺔ ﻋﺪﻡ ﻣﻌﺮﻓﺔ ﻗﻴﻤﺔ ﺍﳌﻌﻠﻤﺔ ﺍﳌﻼﺋﻤﺔ ‪ ، a‬ﺇﺿﺎﻓﺔ ﺇﱃ ‪ . yT‬ﺇﻻ ﺃﻧﻪ ﻭﺑﺴﺒﺐ ﻏﻴﺎﺏ‬
‫ﻃﺮﻳﻘﺔ ﺗﻘﺪﻳﺮ ﻟﻠﻤﻌﻠﻤﺎﺕ ﰲ ﺍﻟﻌﻤﻠﻴﺔ ﻫﺬﻩ‪ ،‬ﻓﺈﻧﻪ ﻻ ﳝﻜﻦ ﺍﳊﻜﻢ ﻋﻠﻰ ﻣﺪﻯ ﳒﺎﺡ‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﺨﺘﺎﺭ ﰲ ﺃﻋﻄﺎﺀ ﻗﻴ‪‬ﻢ ﻟﻠﻤﺘﻐﲑ ﺍﳌﻌﲏ ﰲ ﺍﻟﻔﺘﺮﺓ ﺍﳌﺪﺭﻭﺳﺔ‪ ،‬ﻭﺍﻟﱵ ﺗﻘﺘﻔﻲ ﺃﺛﺮ‬
‫ﻣﺸﺎﻫﺪﺍﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ‪ ،‬ﻧﺘﻴﺠﺔ ﻏﻴﺎﺏ ﺍﻷﺩﻭﺍﺕ ﺍﻹﺣﺼﺎﺋﻴﺔ ﺍﻻﺧﺘﺒﺎﺭﻳﺔ‬
‫ﺍﻟﻀﺮﻭﺭﻳﺔ ﻟﻠﻌﻤﻠﻴﺔ ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﺳﺎﺑﻘﺎﹰ‪ .‬ﳍﺬﺍ ﺍﻟﻐﺮﺽ ﺃﺭﺩﻧﺎ ﺗﻘﺪﱘ ﻃﺮﻕ ﲤﻬﻴﺪ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻟﻸﺳﺒﺎﺏ ﺍﻟﺘﺎﻟﻴﺔ ﻭﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻌﻼﻗﺘﲔ ﺍﳌﺬﻛﻮﺭﺗﲔ ﺳﺎﺑﻘﺎ‪.‬‬
‫‪ -1‬ﲤﻬﻴﺪ ﺍﻟﺴﻠﺴﻠﺔ ﺩﺍﺧﻞ ﺍﻟﻌﻴﻨﺔ )‪ (t = 1,2,...,T‬ﳌﻌﺮﻓﺔ ﻣﺪﻯ ﻗﺪﺭﺓ ﺍﻟﻨﻤﻮﺫﺝ‬
‫ﻋﻠﻰ ﺍﻗﺘﻔﺎﺀ ﺃﺛﺮ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ )‪.(Tracking performance‬‬
‫‪ -2‬ﺗﻘﺪﻳﺮ ﺍﳌﻌﻠﻤﺔ ‪ a‬ﺍﻟﱵ ﺗﻀﻤﻦ ﺗﺪﻧﻴﺔ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‪.‬‬
‫‪ -3‬ﺗﺴﻤﺢ ﻫﺬﻩ ﺍﻟﻌﻼﻗﺔ ﺑﺎﻧﻄﻼﻕ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﻣﻦ ﺧﻼﻝ ﺗﻮﻓﲑ ‪ ، ŷT‬ﻭﺍﻟﱵ‬
‫ﻧﺴﺘﻌﻤﻠﻬﺎ ﰲ ﺍﻟﻌﻼﻗﺔ )‪ (3.31‬ﺑﺪﻝ ‪. yT‬‬

‫‪Smoothing‬‬ ‫ﺏ ‪ -‬ﺍﻟﺘﻤﻬﻴﺪ‬
‫ﻧﻔﺮﻕ ﺑﲔ ﻋﻤﻠﻴﺘﲔ‪ ،‬ﺍﻷﻭﱃ ﺗﺘﻤﺜﻞ ﰲ ﺍﻟﺘﻤﻬﻴﺪ‪ ،‬ﺍﻟﺬﻱ ﻧﻌﲏ ﺑﻪ ‪‬ﺬﻳﺐ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻣﻦ ﺧﻼﻝ ﺇﺯﺍﻟﺔ ﺍﳊﻮﺍﺩﺙ ﺍﻟﻌﺎﺭﺿﺔ )ﺍﻟﺘﺬﺑﺬﺑﺎﺕ ﺍﳊﺎﺩﺓ‬
‫ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ( ﻋﻨﻬﺎ ﻟﺘﺴﻬﻴﻞ ﻋﻤﻠﻴﺔ ﺍﻟﺘﺤﻠﻴﻞ ﻭﺍﻟﺘﻔﺴﲑ‪ .‬ﺳﻨﺮﻯ ﻓﻴﻤﺎ ﺑﻌﺪ ﺃﻥ ﻫﺬﻩ‬
‫ﺍﻟﺘﻘﻨﻴﺔ ﺗﻌﺘﻤﺪ ﻛﺘﻨﺒﺆ ﺗﺎﺭﳜﻲ ﺍﺧﺘﺒﺎﺭﻱ ﺩﺍﺧﻞ ﺍﻟﻌﻴﻨﺔ ﰲ ﺑﺪﺍﻳﺔ ﺍﻷﻣﺮ‪ ،‬ﺑﻴﻨﻤﺎ ﺍﳋﻄﻮﺓ‬
‫ﺍﻟﺜﺎﻧﻴﺔ ﺗﻨﻌﻜﺲ ﰲ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﺍﻟﺬﻱ ﻳﻜﻮﻥ ﺗﻨﺒﺆﺍﹰ ﻋﻤﻠﻴﺎﹰ ﺧﺎﺭﺝ ﺍﻟﻌﻴ‪‬ﻨﺔ‪ ،‬ﻳ‪‬ﻌﺘﻤﺪ ﻋﻨﺪ‬

‫‪88‬‬
‫ﲡﺎﻭﺯ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﳋﺎﺻﺔ ﺑﺎﺧﺘﻴﺎﺭ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻼﺋﻢ‪ .‬ﻫﺬﺍ ﺍﻟﺘﻨﺒﺆ ﳝﻜﻦ ﺍﻻﻋﺘﻤﺎﺩ‬
‫ﻋﻠﻴﻪ ﰲ ﻋﻤﻠﻴﺔ ﺍﲣﺎﺫ ﺍﻟﻘﺮﺍﺭ‪.‬‬

‫ﻃﺮﻕ ﲤﻬﻴﺪ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪:‬‬

‫ﺑﺪﺍﻳﺔ ﻧﺴﺘﻌﲔ ﺑﺘﻘﻨﻴﺔ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﰲ ﳐﺘﻠﻒ ﺃﺷﻜﺎﳍﺎ ﺍﻟﺒﺴﻴﻄﺔ‬
‫ﻟﺘﻤﻬﻴﺪ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﺍﻟﺒﺤﺘﺔ ﺇﺿﺎﻓﺔ ﻟﺒﺘﻘﻨﻴﺎﺕ ﺍﻟﺘﻤﻬﻴﺪ ﺍﻷﺳﻲ‪.‬‬
‫‪N-period moving averages‬‬ ‫ﺍ‪ -‬ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﻟﺒﺴﻴﻄﺔ‪:‬‬
‫ﻫﻲ ﺃﺑﺴﻂ ﻃﺮﻳﻘﺔ ﻧﺘﻨﺎﻭﳍﺎ ﺍﻵﻥ ﻭﻫﻲ ﻣﻌﻄﺎﺓ ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫~‬ ‫‪1‬‬
‫= ‪yt‬‬ ‫) ‪( y t + y t - 1 + y t - 2 +... y t - n + 1‬‬
‫‪n‬‬
‫~‬ ‫‪1 n -1‬‬ ‫)‪(3.32‬‬
‫= ‪yt‬‬ ‫‪å y t-r‬‬
‫‪n r= 0‬‬
‫ﻣﺜﺎﻝ ‪ :16‬ﻟﺪﻳﻨﺎ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ ﻭﺍﳌﻄﻠﻮﺏ ﲤﻬﻴﺪ ﺍﻟﺴﻠﺴﻠﺔ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺘﻮﺳﻄﺎﺕ‬
‫ﺍﳌﺘﺤﺮﻛﺔ ﳌﺎ ) ‪. ( n = 4‬‬

‫ﺍﳉﺪﻭﻝ )‪ :(20‬ﻣﻌﻄﻴﺎﺕ ﺍﻗﺘﺼﺎﺩﻳﺔ‬

‫‪10‬‬ ‫‪9‬‬ ‫‪8‬‬ ‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪t‬‬
‫‪81‬‬ ‫‪72‬‬ ‫‪87‬‬ ‫‪102‬‬ ‫‪75‬‬ ‫‪96‬‬ ‫‪84‬‬ ‫‪66‬‬ ‫‪60‬‬ ‫‪78‬‬ ‫‪y‬‬

‫ﺇﻥ ﺗﻄﺒﻴﻖ ﺍﻟﻌﻼﻗﺔ )‪ (3.31‬ﺃﻋﻼﻩ ﻳﻘﺘﻀﻲ ﺑﺪﺍﻳﺔ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻤﻬﻴﺪ ﻣﻦ ﺍﻟﻔﺘﺮﺓ‬

‫ﺍﻟﺮﺍﺑﻌﺔ ﻟﺘﻔﺎﺩﻱ ﺍﻟﺴﻘﻮﻁ ﰲ ﺇﺷﻜﺎﻟﻴﺔ ﻋﺪﻡ ﺗﻮﻓﺮ ﻣﺸﺎﻫﺪﺍﺕ ﻗﺒﻞ ﺗﻠﻚ ﺍﻟﻔﺘﺮﺓ ﻭﺍﻟﱵ‬
‫ﻧﻄﻠﻖ ﻋﻠﻴﻬﺎ ﻗﻴﻢ ﺍﻻﻧﻄﻼﻕ‪ ،‬ﺇﺫﺍﹰ‪:‬‬
‫~‬ ‫‪1‬‬
‫= ‪y4‬‬ ‫) ‪( y4 + y3 + y2 + y1‬‬
‫‪4‬‬
‫‪1‬‬
‫=‬ ‫) ‪( 84 + 66 + 60 + 78‬‬
‫‪4‬‬

‫‪89‬‬
 ‫~‬ ‫‪1‬‬
‫= ‪y5‬‬ ‫) ‪( y5 + y4 + y3 + y2‬‬
‫‪4‬‬

‫‪1‬‬
‫=‬ ‫) ‪( 96 + 84 + 66 + 60‬‬
‫‪4‬‬
‫‪...‬‬
‫~‬ ‫‪1‬‬
‫= ‪y 10‬‬ ‫) ‪( y 10 + y 9 + y 8 + y 7‬‬
‫‪4‬‬

‫‪1‬‬
‫=‬ ‫) ‪( 81 + 72 + 87+ 102‬‬
‫‪4‬‬

‫ﻣﻦ ﺧﻼﻝ ﻫﺬﺍ ﺍﳌﺜﺎﻝ ﻧﻼﺣﻆ ﻣﺎ ﻳﻠﻲ‪:‬‬

‫ﻣﻦ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﻷﻭﱃ‪ ،‬ﳓﺘﺎﺝ ﺇﱃ‬ ‫)‪ (3.32‬ﻋﻠﻰ ) ‪( n - 1‬‬
‫‪ -1‬ﺃﻥ ﺗﻄﺒﻴﻖ ﺍﻟﻌﻼﻗﺔ‬
‫ﻗﻴ‪‬ﻢ ﺍﺑﺘﺪﺍﺋﻴﺔ ﻧﺴﻤﻴﻬﺎ ﻗﻴ‪‬ﻢ ﺍﻻﻧﻄﻼﻕ ﺣﻴﺚ‪:‬‬
‫ﺍ‪ -‬ﺇﺫﺍ ﺗﻮﻓﺮﺕ ﳎﻤﻮﻋﺔ ﻣﻌﺘﱪﺓ ﻣﻦ ﺍﳌﺸﺎﻫﺪﺍﺕ‪ ،‬ﻧﺴﺘﻄﻴﻊ ﺇﳘﺎﻝ ﺍﻝ‬
‫) ‪ ( n - 1‬ﻣﻦ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﻷﻭﱃ‪.‬‬
‫ﺑـ‪ -‬ﰲ ﺍﳊﺎﻟﺔ ﺍﻷﺧﺮﻯ ﻧﻐﻴ‪‬ﺮ ﻣﻦ ﺍﻟﻘﺎﻋﺪﺓ ﻛﺄﻥ ﳓﺴﺐ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ‬
‫ﻟﻠﻤﺸﺎﻫﺪﺍﺕ ﺍﻷﻭﱃ ﺍﻟﱵ ﺗ‪‬ﻘﺤﻢ ﰲ ﻋﻤﻠﻴﺔ ﺍﳊﺴﺎﺏ‪.‬‬

‫‪ -2‬ﺍﺧﺘﻴﺎﺭ ) ‪ ( n‬ﻛﻤﺎ ﺫﻛﺮ ﺳﺎﺑﻘﹰﺎ ﳝﻜﻦ ﻃﺮﺡ ﺍﳋﻴﺎﺭﺍﺕ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪n =4‬‬ ‫ﺃ‪ -‬ﰲ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﻔﺼﻠﻴﺔ ﻧﻔﻀﻞ ﺃﻥ ﺗﻜﻮﻥ‬
‫ﺏ‪ -‬ﰲ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺸﻬﺮﻳﺔ ﻧﻔﻀﻞ ﺃﻥ ﺗﻜﻮﻥ ‪n = 12‬‬
‫ﺕ‪ -‬ﰲ ﺍﳊﺎﻻﺕ ﺍﻷﺧﺮﻯ ﳔﺘﺎﺭﻫﺎ ﺑﻄﺮﻳﻘﺔ ﻋﻔﻮﻳﺔ‪.‬‬
‫ﺭﻏﻢ ﺫﻟﻚ ﻟﻘﺪ ﺍﻧﺘﻘﺪﺕ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻋﻠﻰ ﺃﺳﺎﺱ ﺃ‪‬ﺎ ﲤﻬﺪ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﺍﻋﺘﻤﺎﺩﺍ ﻋﻠﻰ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﳌﺎﺿﻴﺔ ﻓﻘﻂ‪ ،‬ﻭﱂ ﺗﺄﺧﺬ ﺍﳌﻼﺣﻈﺎﺕ ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﺑﻌﲔ‬
‫ﺍﻻﻋﺘﺒﺎﺭ‪.‬‬

‫‪90‬‬
 ‫‪Centered Moving Averages‬‬ ‫‪ -2‬ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﳌﻤﺮﻛﺰﺓ‪:‬‬
‫ﺗﺄﺧﺬ ﻫﺬﻩ ﺍﻟﺘﻘﻨﻴﺔ ﺍﻻﻧﺘﻘﺎﺩ ﺍﻟﺴﺎﺑﻖ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ ﻭﺗﻘﺘﺮﺡ ﺍﻟﺘﻘﻨﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫)‪:(odd‬‬ ‫ﺍ‪ -‬ﺣﺎﻟﺔ ‪ n‬ﻓﺮﺩﻳﺔ‬
‫‪n -1‬‬
‫(‬ ‫)‬
‫~‬ ‫‪1‬‬ ‫‪2‬‬
‫= ‪yt‬‬
‫‪n‬‬
‫‪åy‬‬
‫‪n -1‬‬
‫‪t-r‬‬ ‫)‪(3.33‬‬
‫(‪r = -‬‬ ‫)‬
‫‪2‬‬
‫ﻓﺈﻥ‪:‬‬ ‫‪n=5‬‬ ‫ﻓﻔﻲ ﺣﺎﻟﺔ‬
‫~‬ ‫‪1‬‬
‫= ‪yt‬‬ ‫) ‪( y t + 2 + y t +1 + y t + y t -1 + y t - 2‬‬
‫‪5‬‬
‫)‪(even‬‬ ‫ﺏ‪ -‬ﺣﺎﻟﺔ ‪ n‬ﺯﻭﺟﻴﺔ‬
‫‪n‬‬
‫) (‬
‫~‬ ‫‪1‬‬ ‫‪2‬‬
‫= ‪yt‬‬ ‫‪å D t y t-r‬‬ ‫)‪(3.34‬‬
‫) ‪n r=-( n‬‬
‫‪2‬‬

‫ﻣﺘﻐﲑ ﲤﺜﻴﻠﻲ ‪ Dummy variable‬ﻳﺄﺧﺬ‬ ‫‪Dt‬‬ ‫ﺣﻴﺚ‬

‫‪ì 1‬‬ ‫‪n ü‬‬
‫‪ï 2 if r = ± 2 ï‬‬
‫‪Dt = í‬‬
‫‪n‬‬ ‫‪ný‬‬
‫‪ï1 if - < r < ï‬‬
‫‪î‬‬ ‫‪2‬‬ ‫‪2þ‬‬

‫ﻓﺈﻥ ﺍﻟﻌﻼﻗﺔ ﺍﻷﺧﲑﺓ ﺗﻜﺘﺐ‪:‬‬ ‫‪n =4‬‬ ‫ﰲ ﺣﺎﻟﺔ‬

‫~‬ ‫‪1 1‬‬
‫= ‪yt‬‬ ‫) ‪( 2 y t + 2 + y t + 1 + y t + y t - 1 + 12 y t - 2‬‬
‫‪4‬‬

‫ﻣﺜﺎﻝ ‪ :17‬ﲤﻬﻴﺪ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺴﺎﺑﻘﺔ ‪‬ﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺇﺫﺍ ﻛﺎﻧﺖ ‪. n = 3‬‬

‫ﺗﻜﺘﺐ ﻫﺬﻩ ﺍﻟﻌﻼﻗﺔ ﻭﻓﻖ ﺍﻟﺸﻜﻞ ﺍﳌﻨﺎﺳﺐ ﺍﻟﺘﺎﱄ‪:‬‬
‫~‬ ‫‪1‬‬
‫= ‪yt‬‬ ‫) ‪( y t+1 + y t + y t -1‬‬
‫‪3‬‬

‫‪91‬‬
 ‫~‬ ‫‪1‬‬
‫= ‪y2‬‬ ‫) ‪( y3 + y2 + y1‬‬
‫‪3‬‬
‫‪1‬‬
‫=‬ ‫) ‪( 66 + 60 + 78‬‬
‫‪3‬‬
‫~‬ ‫‪1‬‬
‫= ‪y3‬‬ ‫) ‪( y4 + y3 + y2‬‬
‫‪3‬‬
‫‪1‬‬
‫=‬ ‫) ‪( 84 + 66 + 60‬‬
‫‪3‬‬
‫‪...‬‬
‫~‬ ‫‪1‬‬
‫= ‪y9‬‬ ‫) ‪( y 10 + y 9 + y 8‬‬
‫‪3‬‬
‫‪1‬‬
‫=‬ ‫) ‪( 81 + 72 + 87‬‬
‫‪3‬‬

‫ﻧﺸﲑ ﺇﱃ ﺃﻥ ﺍﻟﻄﺮﻳﻘﺘﲔ ﺍﻟﺴﺎﺑﻘﺘﲔ‪ ،‬ﻭﺍﳌﺬﻛﻮﺭﺗﲔ ﺁﻧﻔﺎﹰ‪ ،‬ﺗﺴﺘﻌﻤﻼﻥ ﻹﺯﺍﻟﺔ‬

‫ﺍﻟﺬﺑﺬﺑﺎﺕ ﺍﳊﺎﺩﺓ )ﺍﻟﻔﺼﻠﻴﺔ( ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬
‫‪ -3‬ﻃﺮﻕ ﺍﻟﺘﻤﻬﻴﺪ ﺍﻷﺳﻲ‪:‬‬
‫)‪(EWMA‬‬ ‫ﻫﺬﻩ ﺍﻟﻄﺮﻕ ﻣﺸﺘﻘﺔﹰ ﻣﻦ ﳕﻮﺫﺝ ﺑﺮﺍﻭﻥ )‪ (Brown‬ﻭﺍﳌﺴﻤﻰ ﺑـ‬
‫ﺍﳌﺬﻛﻮﺭ ﺳﺎﺑﻘﺎﹰ‪ ،‬ﻭﺍﻟﺬﻱ ﻗﻠﻨﺎ ﺁﻧﺬﺍﻙ ﺃﻧﻪ ﻳﻌﻄﻲ ﻭﺯﻥ ﺃﻛﱪ ﻟﻠﻘﻴ‪‬ﻢ ﺍﳊﺪﻳﺜﺔ ﺯﻣﻨﻴﺎ ﻋﻦ‬
‫ﺳﺎﺑﻘﺘﻬﺎ‪ ،‬ﺃﻳﻦ ﺗﺴﺘﻌﻤﻞ ﻫﺬﻩ ﺍﻟﻄﺮﻕ ﰲ ﻋﻤﻠﻴﺎﺕ ﺍﻟﺘﻨﺒﺆ ﺍﳋﺎﺻﺔ ﺑﺎﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬

‫ﺍﻷﺣﺎﺩﻱ‪Single Exponential Smoothing :‬‬ ‫‪ -1-3‬ﳕﻮﺫﺝ ﺍﻟﺘﻤﻬﻴﺪ ﺍﻷﺳﻲ‬

‫ﺗﻜﻮﻥ ﻫﺬﻩ ﺍﻟﺘﻘﻨﻴﺔ ﻗﺎﺑﻠﺔ ﺍﻻﺳﺘﻌﻤﺎﻝ ﰲ ﺣﺎﻟﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ‬
‫ﺍﳌﺘﺬﺑﺬﺑﺔ ﺣﻮﻝ ﻭﺳﻂ ﺣﺴﺎﰊ ﺛﺎﺑﺖ‪ .‬ﻳﻌﻄﻰ ﺍﻟﻨﻤﻮﺫﺝ ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫~‬
‫‪y‬‬ ‫‪t‬‬ ‫‪= ay‬‬ ‫‪t‬‬ ‫‪+ a( 1 - a )y‬‬ ‫‪t-1‬‬ ‫‪+ a( 1 - a )2 y‬‬ ‫‪t- 2‬‬ ‫‪+‬‬

‫‪a( 1 - a )3 y‬‬ ‫‪t- 3‬‬ ‫) ‪+... + a ( 1 - a‬‬ ‫‪m‬‬

‫‪y‬‬ ‫‪t-m‬‬ ‫‪...‬‬
‫~‬ ‫‪¥‬‬
‫‪y t = aå ( 1 - a ) r y t-r‬‬ ‫)‪(3.35‬‬
‫‪r= 0‬‬

‫‪92‬‬
 ‫ﻧﻈﺮﺍﹰ ﻟﻌﻴﻮ‪‬ﺎ ﻧﻮﺩ ﺗﺒﺴﻴﻄﻬﺎ ‪‬ﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ‬
‫ﺍ ‪ -‬ﻧﺆﺧﺮ ﺍﳌﻌﺎﺩﻟﺔ ﺍﳌﻔﻜﻜﺔ ﺃﻋﻼﻩ ﺑﻔﺘﺮﺓ ﺯﻣﻨﻴﺔ ﻭﺍﺣﺪﺓ‪ ،‬ﰒ ﻧﻀﺮ‪‬ﺎ ﰲ‬
‫ﺍﳌﻘﺪﺍﺭ ) ‪.( 1 - a‬‬
‫~‬
‫‪2‬‬ ‫‪3‬‬
‫‪(1 - a ) y t -1 = a (1 - a ) yt -1 + a (1 - a ) yt -2 + a (1 - a ) yt -3 + .......‬‬
‫‪m+1‬‬
‫) ‪+ a (1 - a‬‬ ‫‪yt -m-1...‬‬

‫ﺑـ ‪ -‬ﻃﺮﺡ ﻫﺬﻩ ﺍﻷﺧﲑﺓ ﻣﻦ ﺍﳌﻌﺎﺩﻟﺔ ﺍﻷﺻﻠﻴﺔ ﳓﺼﻞ ﻋﻠﻰ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫~‬ ‫~‬
‫‪y t - ( 1 - a ) y t - 1 = ay t‬‬

‫ﺟـ‪ -‬ﺑﺈﻋﺎﺩﺓ ﺍﻟﺘﺮﺗﻴﺐ‪ ،‬ﻧﻜﺘﺐ ﺍﻟﻌﻼﻗﺔ ﺍﻟﻨﻬﺎﺋﻴﺔ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﻌﻤﻠﻲ ﺍﻟﺘﺎﱄ‪:‬‬

‫~‬ ‫~‬
‫‪y t = ay t + ( 1 - a ) y t - 1‬‬ ‫)‪(3.36‬‬

‫)‪(Starting values‬‬ ‫ﺣﻴﺚ ‪ t = 2, T‬ﻭﻣﻨﻪ ﳓﺘﺎﺝ ﺇﱃ ﻗﻴ‪‬ﻢ ﺍﺑﺘﺪﺍﺋﻴﺔ‬

‫ﻻﻧﻄﻼﻕ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻤﻬﻴﺪ ﻟﺬﺍ ﻧﻔﺘﺮﺽ ﺃﻥ ‪ ~y1 = y1‬ﻛﻤﺎ ﻳﻮﺿﺤﻪ ﺍﳌﺜﺎﻝ ‪.13‬‬
‫ﻳﻜﺘﺐ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺘﻨﺒﺆﻱ ﺍﳌﺴﺘﻘﺒﻠﻲ ﻟﻠﻔﺘﺮﺓ ‪ T+1‬ﺇﱃ ‪ T+L‬ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫~ ) ‪yˆT +1 = a yT + (1 - a‬‬
‫‪yT‬‬ ‫)‪(3.37‬‬

‫)‬
‫ﺃﻳﻦ ﻧﻌﻮﺽ ﺃﺛﻨﺎﺀ ﺍﻟﺘﻨﺒﺆ ‪ yT‬ﰲ ﺍﻟﻌﻼﻗﺔ )‪ (3.31‬ﺑـ ‪ ~yT‬ﻭ ﻣﻨﻪ‬
‫‪yˆT +L = ... = yˆT +1‬‬

‫ﻧﻜﺮﺭ ﺍﻹﺷﺎﺭﺓ ﺇﱃ ﻭﺟﻮﺏ ﻛﻮﻥ ﺍﻟﻔﺘﺮﺓ ‪ L‬ﻗﺼﲑﺓ ﺟﺪﺍﹰ ﺣﱴ ﻻ ﻳﻔﻘﺪ ﺍﻟﺘﻨﺒﺆ‬

‫ﻣﺼﺪﺍﻗﻴﺘﻪ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﻳﺴﻤﺢ ﻟﻠﻤﺴﻴ‪‬ﺮ ﻣﺜﻼ ﺍﲣﺎﺫ ﻗﺮﺍﺭﺍﺗﻪ ﺑﺪﺭﺟﺔ ﻭﺍﻓﻴﺔ ﻣﻦ ﺍﻟﺮﺍﺣﺔ‬
‫ﻭﺍﳊﺮﻳﺔ‪.‬‬

‫‪93‬‬
‫ﻣﺜﺎﻝ ‪ :18‬ﻟﺪﻳﻨﺎ ﺍﻟﺴﻠﺴﻠﺔ ‪] y t‬ﺍﳉﺪﻭﻝ)‪ [ (21‬ﺍﳌﻤﺜﻠﺔ ﻟﺴﻌﺮ ﺍﻟﱪﺍﻧﺖ )‪ (Brent‬ﰲ‬
‫ﺃﺣﺪ ﺍﻷﺳﻮﺍﻕ ﺍﻟﺪﻭﻟﻴﺔ ﻣﻦ ‪ ،1994.12-1993.07‬ﻭ ‪ ، a = 0.95‬ﺃﻳﻦ‬
‫~‬
‫‪y 1 = y 1 = 16 . 76‬‬
‫ﻫﻲ ﲤﺜﻞ ﻗﻴﻤﺔ ﺍﻻﻧﻄﻼﻕ )ﺍﻻﺑﺘﺪﺍﺋﻴﺔ(‪ .‬ﻭﻓﻖ ﻫﺬﻩ ﺍﻟﺴﲑﻭﺭﺓ ﻭﺑﻌﺪ ﲢﺪﻳﺪ ﻗﻴﻢ‬
‫ﺍﻻﻧﻄﻼﻕ‪ ،‬ﳝﻜﻦ ﺣﺴﺎﺏ ﻣﻜﻮﻧﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﻤﻬﺪﺓ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫~‬ ‫~‬
‫) ‪y 2 = 0. 95 ( y 2 ) + ( 1 - 0. 95 )( y 1‬‬
‫~‬
‫) ‪y 2 = 0. 95( 16. 71 ) + ( 1 - 0. 95 )( 16. 76‬‬
‫‪= 16. 7125‬‬
‫~‬ ‫~‬
‫) ‪y 3 = 0. 95( y 3 ) + ( 1 - 0. 95 )( y 2‬‬
‫~‬
‫) ‪y 3 = 0. 95( 16. 03 ) + ( 1 - 0. 95 )( 16. 7125‬‬
‫‪= 16. 064‬‬
‫ﻭﻫﻜﺬﺍ ﺇﱃ ﻏﺎﻳﺔ ﺍﳌﺸﺎﻫﺪﺓ ﺍﳌﺘﻮﻓﺮﺓ ﺍﻷﺧﲑﺓ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫~‬ ‫~‬ ‫~‬
‫) ‪y T = y 18 = 0. 95( y 18 ) + ( 1 - 0. 95 )( y 17‬‬
‫~‬
‫) ‪y 18 = 0. 95( 19. 25 ) + ( 1 - 0. 95 )( 20. 863‬‬
‫‪= 19. 33‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(21‬ﺍﳌﻌﻄﻴﺎﺕ ﺍﻷﺻﻠﻴﺔ ﻭﺍﳌﻤﻬﺪﺓ‬
‫‪t‬‬ ‫‪yt‬‬ ‫~‬
‫‪yt‬‬ ‫‪t‬‬ ‫‪y‬‬ ‫~‬
‫‪yt‬‬
‫‪t‬‬

‫‪1‬‬ ‫‪16.76‬‬ ‫‪16.76‬‬ ‫‪12‬‬ ‫‪21.77‬‬ ‫‪21.7567‬‬

‫‪2‬‬ ‫‪16.71‬‬ ‫‪16.7125‬‬ ‫‪13‬‬ ‫‪22.27‬‬ ‫‪22.2443‬‬
‫‪3‬‬ ‫‪16.03‬‬ ‫‪16.0641‬‬ ‫‪14‬‬ ‫‪24.08‬‬ ‫‪23.9882‬‬
‫‪4‬‬ ‫‪16.53‬‬ ‫‪16.5067‬‬ ‫‪15‬‬ ‫‪20.59‬‬ ‫‪20.7599‬‬
‫‪5‬‬ ‫‪15.10‬‬ ‫‪15.1703‬‬ ‫‪16‬‬ ‫‪20.73‬‬ ‫‪20.7314‬‬
‫‪6‬‬ ‫‪13.54‬‬ ‫‪13.6215‬‬ ‫‪17‬‬ ‫‪20.87‬‬ ‫‪20.863‬‬
‫‪7‬‬ ‫‪17.61‬‬ ‫‪17.4105‬‬ ‫‪18‬‬ ‫‪19.25‬‬ ‫‪19.3306‬‬
‫‪8‬‬ ‫‪18.52‬‬ ‫‪18.4645‬‬ ‫‪19‬‬ ‫‪-‬‬ ‫‪19.254‬‬
‫‪9‬‬ ‫‪18.01‬‬ ‫‪18.0327‬‬ ‫‪20‬‬ ‫‪-‬‬ ‫‪19.254‬‬
‫‪10‬‬ ‫‪20.17‬‬ ‫‪20.0631‬‬ ‫‪21‬‬ ‫‪-‬‬ ‫‪19.254‬‬
‫‪11‬‬ ‫‪21.58‬‬ ‫‪21.5041‬‬ ‫‪22‬‬ ‫‪-‬‬ ‫‪19.254‬‬

‫‪94‬‬
 ‫ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﻭﻓﻖ ﺍﳌﺮﺣﻠﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫) ‪$y T+1 = a ( y 18 ) + ( 1 - a )( $y 18‬‬
‫~‬
‫‪$y T+1 = 19. 254‬‬ ‫ﻭﻣﻨﻪ‪،‬‬ ‫‪$y T = y T‬‬ ‫ﺃﻳﻦ‬
‫‪$y T+2 = 19. 254‬‬ ‫ﺑﺎﻟﺘﺸﺎﺑﻪ‪:‬‬
‫‪...‬‬
‫‪$y T+ L = 19. 254‬‬

‫ﺍﻟﺸﻜﻞ )‪ :(19‬ﻣﻨﺤﲎ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ ﻭﺍﳌﻤﻬﺪﺓ ﺍﳌﺘﻄﺎﺑﻘﲔ ﺗﻘﺮﻳﺒﺎﹰ‬

‫‪Double Exponential Smoothing‬‬ ‫‪ -2-3‬ﳕﻮﺫﺝ ﺍﻟﺘﻤﻬﻴﺪ ﺍﻷﺳﻲ ﺍﻟﺜﻨﺎﺋﻲ‪:‬‬

‫ﻓﺈﺫﺍ ﻛﺎﻧﺖ ﺍﻟﺴﻠﺴﻠﺔ ﲢﻮﻱ ﺇﺿﺎﻓﺔ ﺇﱄ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻣﺮﻛﺒﺔ ﺍﲡﺎﻩ ﻋﺎﻡ‪،‬‬
‫ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻨﻬﺎ ﻭﻓﻖ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻻﳓﺪﺍﺭﻳﺔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y t = b 0 + b1 t + u t‬‬

‫ﺃﻳﻦ ‪ b 0 + b 1 t‬ﲤﺜﻞ ﻣﺮﻛﺒﺔ ﺍﲡﺎﻩ ﻋﺎﻡ ﺧﻄﻲ ﻭ ‪ ut‬ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻭ ﺍﻟﱵ ﳝﻜﻦ‬

‫ﲤﻬﻴﺪﻫﺎ ‪‬ﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻭﻓﻖ ﺍﳌﺮﺣﻠﺘﲔ ﺍﻟﺘﺎﻟﻴﺘﲔ‪:‬‬

‫‪95‬‬
 ‫ﺍﳌﺮﺣﻠﺔ ﺍﻷﻭﱃ‪:‬‬
‫‪.‬‬ ‫~‬
‫~)‪yt = ayt + (1-a‬‬
‫‪yt-1‬‬ ‫)‪(3.38‬‬

‫ﺍﳌﺮﺣﻠﺔ ﺍﻟﺜﺎﻧﻴﺔ‪:‬‬
‫»‬ ‫~‬ ‫»‬
‫‪y t = a y t + (1 - a ) y t -1‬‬ ‫)‪(3.39‬‬

‫ﻳﺘﻢ ﺣﺴﺎﺏ ﺍﳌﻌﻠﻤﺘﲔ ﻛﻤﺎ ﻳﻠﻲ ‪:‬‬

‫‪8‬‬

‫~‬ ‫»‬
‫‪b0 = 2 y T - y T‬‬ ‫)‪(3.40‬‬

‫‪a‬‬ ‫~‬ ‫»‬

‫)‪(3.41‬‬
‫= ‪b1‬‬ ‫) ‪( yT- yT‬‬
‫‪1-a‬‬

‫ﻳﺄﺧﺬ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺘﻨﺒﺌﻲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪yˆT+L = b0 + b1l‬‬ ‫)‪(3.42‬‬

‫‪.l=1,L‬‬ ‫ﺣﻴﺚ ‪ l‬ﳝﺜﻞ ﺃﻓﻖ ﺍﻟﺘﻨﺒﺆ ﺃﻱ‬

‫ﺍﳌﻌﻠﻤﺘﲔ‪:Holt Two Smoothing Parameters‬‬ ‫‪ -3-3‬ﻃﺮﻳﻘﺔ ﻫﻮﻟﺖ ﺫﺍﺕ‬
‫ﻳ‪‬ﻠﺠﺄ ﺇﱃ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﰲ ﻧﻔﺲ ﻇﺮﻭﻑ ﺍﺳﺘﻌﻤﺎﻝ ﺍﻟﺘﻘﻨﻴﺔ ﺍﻟﺴﺎﺑﻘﺔ‪ ،‬ﻭﻫﺬﺍ‬
‫ﻃﺒﻌ‪‬ﺎ ﻻ ﻳﻌﲏ ﺃ‪‬ﺎ ﺗﻌﻄﻲ ﻧﻔﺲ ﺍﻟﻨﺘﺎﺋﺞ ‪ .‬ﺗﺘﻜﻮﻥ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻣﻦ ﻣﻌﺎﺩﻟﺘﲔ ﻭﻛﺬﺍ‬
‫ﺛﺎﺑﱵ ﲤﻬﻴﺪ ﺃﺣﺪﳘﺎ ﺧﺎﺹ ﺑﺎﻟﻌﺸﻮﺍﺋﻴﺔ ﻭﺍﻵﺧﺮ ﺑﺎﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﻭﺗﻜﺘﺐ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫~‬ ‫~‬
‫) ‪y t = ay t + ( 1 - a ) ( y t -1 + rt -1‬‬ ‫)‪(3.43‬‬

‫~‬ ‫~‬
‫‪rt = g ( y t - y t -1 ) + ( 1 - g ) rt -1‬‬ ‫)‪(3.44‬‬

‫‪P.153 Guy Ansion, Les Méthodes de Prévision en Economie,:‬‬ ‫‪ -8‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ‬

‫‪96‬‬
‫ﻟﻠﺘﺨﻠﺺ ﻣﻦ ﺇﺷﻜﺎﻟﻴﺔ ﻗﻴ‪‬ﻢ ﺍﻻﻧﻄﻼﻕ‪ ،‬ﻧﻘﺘﺮﺡ ﻣﻦ ﺑﲔ ﳎﻤﻮﻋﺔ ﻣﻦ ﺍﻟﺼﻴ‪‬ﻎ‪،‬‬
‫ﺍﻟﺼﻴﻐﺘﲔ ﺍﻟﺘﺎﻟﻴﺘﲔ‪:‬‬
‫~‬
‫‪y1 = y 1‬‬

‫‪r1 = 0‬‬
‫ﺃﻭ‬
‫~‬
‫‪y2 = y 2‬‬

‫‪r2 = y 2 - y 1‬‬

‫‪‬ﺬﺍ ﺗﻨﻄﻠﻖ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻤﻬﻴﺪ ﻣﻦ ﺍﻟﻔﺘﺮﺓ ﺍﻟﺜﺎﻧﻴﺔ ﰲ ﺍﳊﺎﻟﺔ ﺍﻷﻭﱃ ﻭﻣﻦ ﺍﻟﺜﺎﻟﺜﺔ ﰲ‬
‫ﺍﳊﺎﻟﺔ ﺍﻟﺜﺎﻧﻴﺔ‪.‬‬
‫ﻷﻏﺮﺍﺽ ﺍﻟﺘﻨﺒﺆ‪ ،‬ﺗﻜﺘﺐ ﺗﻠﻚ ﺍﳌﻌﺎﺩﻟﺘﲔ ﰲ ﺍﻟﺼﻴﻐﺔ ﺍﳌﻌﺪ‪‬ﻟﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫^‬ ‫~‬
‫‪y T + L = y T + lrT‬‬ ‫)‪(3.45‬‬

‫‪rT = ~y T - ~y T- 1‬‬ ‫ﻫﻮ ﺗﻘﺮﻳﺒﺎﹰ ﻧﻔﺲ ﳕﻮﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﶈﻠﻲ )‪ (3.9‬ﺣﻴﺚ‬
‫ﳌﺎ ‪ g = 1‬ﰲ ﺍﻟﻌﻼﻗﺔ )‪.(3.44‬‬

‫‪Decomposition method‬‬ ‫‪ -4-3‬ﻃﺮﻳﻘﺔ ﺍﻟﺘﻔﻜﻴﻚ‪:‬‬

‫ﺗﺘﻤﺜﻞ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﰲ ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﻣﻦ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﰲ ﺇﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ‬
‫ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺑﻄﺮﻳﻘﺔ ﻣﻼﺋﻤﺔ ﰒ ﲤﻬﻴﺪ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻨﺎﲡﺔ ﻭﺍﳋﺎﻟﻴﺔ‬
‫ﻣﻦ ﺍﳌﺮﻛﺒﺔ ﺍﳌﻨﺰﻭﻋﺔ ﺑﻄﺮﻳﻘﺔ ﺍﻟﺘﻤﻬﻴﺪ ﺍﻵﺳﻲ ﺍﻷﺣﺎﺩﻱ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ‪ ،‬ﰒﹶ ﺍﻟﻘﻴﺎﻡ‬
‫ﺑﻌﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﺑﻄﺮﻳﻘﺔ ﻋﻜﺴﻴﺔ‪.‬‬

‫‪97‬‬
 ‫ﻃﺮﻕ ﺇﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪:‬‬
‫ﻧﺴﺘﻌﺮﺽ ﺍﻟﻄﺮﻕ ﺍﳌﻨﺎﺳﺒﺔ ﻹﺑﻌﺎﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﳋﻄﻲ ﺧﺎﺻﺔ ﻣﻦ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻛﻤﺎ ﳝﻜﻦ ﺍﺳﺘﻌﻤﺎﳍﺎ ﻭﺑﺘﻌﺪﻳﻞﹴ ﺑﺴﻴﻂ ﻣﻊ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻏﲑ‬
‫ﺍﳋﻄﻲ‪.‬‬

‫ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‪:‬‬ ‫‪-1‬‬

‫ﺗﺘﻢ ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ ﺑﺘﻄﺒﻴﻖ ﺍﳌﻌﺎﺩﻟﺔ )‪ (2.6‬ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬

‫‪D y t = y y - y t -1‬‬

‫ﺳﻠﺴﻠﺔ ﺧﺎﻟﻴﺔ ﻣﻦ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬ ‫‪D yt‬‬ ‫ﺃﻳﻦ ﺗﺼﺒﺢ‬

‫ﻣﺜﺎﻝ ‪ :19‬ﺇﺫﺍ ﻛﺎﻧﺖ ﻟﺪﻳﻨﺎ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﻌﺒ‪‬ﺮ ﻋﻨﻬﺎ ﺑﺎﳌﻌﺎﺩﻟﺔ ﺍﻟﺘﺎﻟﻴﺔ ﻭﰲ ﻋﻼﻗﺔ‬
‫ﲡﻤﻴﻌﻴﺔ‪ ،‬ﺍﳌﻄﻠﻮﺏ ﺣﺴﺎﺏ ﺭﻳﺎﺿﻴﺎ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫‪y t = a + bx t + g t + u t‬‬
‫ﺃﻳﻦ ﺗﺘﻤﺜﻞ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﳌﻘﺪﺭﺓ ﰲ ﺍﳌﻘﺪﺍﺭ‪:‬‬
‫‪a+gt‬‬
‫ﺑﺘﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﳓﺼﻞ ﻋﻠﻰ‪:‬‬
‫‪D y t = w t = y y - y t -1‬‬
‫‪y t -1 = a + bx t -1 + g ( t - 1 ) + u t -1‬‬ ‫ﺃﻳﻦ‬
‫ﺇﺫﺍﹰ‬
‫) ‪w t = b ( x t - x t -1 ) + g ( t - ( t - 1 ) ) + ( u t - u t - 1‬‬

‫‪w t = b( x t - x t -1 ) + g + v t‬‬

‫‪w t = g + bz t + v t‬‬

‫‪98‬‬
 ‫ﻫﻲ ﺳﻠﺴﻠﺔ ﺧﺎﻟﻴﺔ ﻣﻦ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺃﻳﻦ‪:‬‬
‫‪zt =xt-x‬‬ ‫‪t -1‬‬

‫‪v t = e t - e t -1‬‬ ‫ﻭ‬

‫ﺍﻟﻄﺮﻳﻘﺔ ﺍﻻﳓﺪﺍﺭﻳﺔ‬ ‫‪-2‬‬

‫ﺇﺫﺍ ﻛﺎﻥ ﻟﺪﻳﻨﺎ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪yt = a+gt + u t‬‬

‫ﺃﻳﻦ ‪ a + g t‬ﲤﺜﻞ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺑﻴﻨﻤﺎ ﲤﺜﻞ ‪ ut‬ﺍﳌﺮﻛﺒﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻭﻧﺮﻳﺪ ﺇﺑﻌﺎﺩ‬
‫ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻣﻨﻬﺎ ‪‬ﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺣﺴﺐ ﺍﳌﺮﺍﺣﻞ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ -1‬ﺗﻘﺪﻳﺮ ﺍﳌﻌﺎﺩﻟﺔ ﺃﻋﻼﻩ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﻭﺍﳊﺼﻮﻝ ﻋﻠﻰ ﺗﻘﺪﻳﺮ‬
‫ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‬
‫‪yˆt = aˆ + gˆ t‬‬
‫ﻣﻨﻪ‬
‫‪yt = aˆ + gˆ t + et‬‬
‫‪ -2‬ﺍﻹﺯﺍﻟﺔ )ﻭﺗﺘﻢ ﺑﻄﺮﺡ ﺗﻘﺪﻳﺮ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ(‬
‫‪yt - aˆ + gˆ t = et‬‬

‫ﺗﻜﻮﻥ ‪ et‬ﺃﻱ ﺍﻟﺒﻮﺍﻗﻲ ﻫﻲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ .‬ﻭﺍﻟﱵ ﳝﻜﻦ‬

‫ﲤﻬﻴﺪﻫﺎ ﺑﻄﺮﻳﻘﺔ ﺍﻟﺘﻤﻬﻴﺪ ﺍﻵﺳﻲ ﺍﻷﺣﺎﺩﻱ ﰒ ﺍﻟﺘﻨﺒﺆ ﲟﺎ ﻳﻠﻲ‪:‬‬
‫)‬
‫~)‪eT +1 = aeT + (1 -a‬‬
‫‪eT‬‬

‫ﻳﺘﻢ ﺍﻟﺘﻨﺒﺆ ﺍﻟﻨﻬﺎﺋﻲ ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺸﺎﻣﻠﺔ ﺍﻟﺘﺎﻟﻴﺔ ﻭﺫﻟﻚ ﺑﺪﻣﺞ ﻣﻦ ﺟﺪﻳﺪ‬
‫ﻟﻠﻤﺮﻛﺒﺘﲔ ﺍﻟﻨﻈﺎﻣﻴﺔ ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪:‬‬
‫‪yˆT+1 =a+‬‬
‫‪ˆ gˆ (T+1)+eˆT+1‬‬

‫‪99‬‬
‫ﻧﺸﲑ ﺇﱃ ﺃﻧﻪ ﻭﻗﺒﻞ ﺗﻄﺒﻴﻖ ﺍﻟﺘﻘﻨﻴﺔ ﺍﳌﻨﺎﺳﺒﺔ ‪ -‬ﻃﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ ﺃﻭ ﺍﻟﻄﺮﻳﻘﺔ‬
‫ﺍﻻﳓﺪﺍﺭﻳﺔ ‪ -‬ﳚﺐ ﻣﺴﺒﻘﺎﹰ ﻣﻌﺮﻓﺔ ﺃﺻﻞ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﱵ‬
‫ﻗﺪ ﺗﻜﻮﻥ ﻧﺎﲡﺔ ﻋﻦ‪:‬‬
‫‪ :(Random‬ﺍﻟﺬﻱ‬ ‫)‪Walk Trending‬‬ ‫ﺍ‪ -‬ﳕﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ‬
‫ﻳﻜﺘﺐ ﻛﻤﺎ ﺳﻨﺮﻯ ﺿﻤﻦ ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪y t = y t -1 + d + e t‬‬ ‫)‪(3.46‬‬

‫ﺣﻴﺚ‬
‫¾ ‪et‬‬
‫¾‬‫‪N‬‬
‫) ‪®( 0, se2‬‬
‫ﻣﻨﻪ‬
‫‪y t - y t -1 = d + e t‬‬
‫)‪:(Time Trending‬‬ ‫ﺏ‪ -‬ﺍﻟﺰﻣﻦ‬
‫ﺍﻟﺬﻱ ﻳﻜﺘﺐ ﰲ ﻧﻔﺲ ﺷﻜﻞ ﺍﻟﻌﻼﻗﺔ )‪ .(25‬ﻧﺸﲑ ﺇﱃ ﺃﻧﻪ ﳝﻜﻦ ﻛﺘﺎﺑﺔ )ﺍ( ﰲ‬
‫ﺷﻜﻞ )ﺏ( ﺃﻱ‪:‬‬

‫‪y 1 = y 0 + d + e1‬‬

‫‪y 2 = y1 + d + e2‬‬

‫‪= ( y 0 + d + e1 ) + d + e 2‬‬
‫‪2‬‬
‫‪= y 0 + d( 2 ) + å e i‬‬
‫‪i =1‬‬

‫‪...‬‬
‫‪t‬‬
‫‪yt = y0 + d (t ) + å e i‬‬
‫‪i =1‬‬

‫‪= y0 + d (t ) + ut‬‬ ‫)‪(3.47‬‬

‫‪100‬‬
‫ﻫﻮ ﻧﻔﺲ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﺬﻛﻮﺭ ﺳﺎﺑﻘﺎﹰ ﺇﺫﺍ ﻭﺿﻌﻨﺎ ‪ y0 = a‬ﻭ‪ d = b‬ﲟﻌﲎ ﺃﻥ‬
‫ﺍﻟﻨﻤﻮﺫﺟﲔ ﻳﺴﻠﻜﺎﻥ ﺳﻠﻮﻛﺎﹰ ﻣﺘﺸﺎ‪‬ﺎﹰ ﻭﻟﻜﻦ ﻃﺮﻳﻘﺔ ﺍﻟﺘﻌﺎﻣﻞ ﻣﻌﻬﻤﺎ ﻗﺪ ﲣﺘﻠﻒ‪،‬‬
‫ﻛﻮﻥ ﺗﻘﺪﻳﺮ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺜﺎﻧﻴﺔ ﲟﺘﻐﲑﻳﻦ ﺍﻟﺘﺎﺑﻊ ﻭﺍﳌﺴﺘﻘﻞ ﻏﲑ ﻣﺴﺘﻘﺮﻳﻦ ‪(non-‬‬
‫)‪ ،stationary‬ﻳﻄﺮﺡ ﻣﺸﺎﻛﻞ ﻛﺒﲑﺓ ﺃﺷﺎﺭ ﺇﻟﻴﻬﺎ ‪ Granger‬و‪ Newbold‬ﺳﻨﺔ‬
‫‪ 1974‬ﲢﺖ ﺍﺳﻢ ‪ Spurious Regression‬ﺃﻱ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺰﺍﺋﻒ ﺃﻭ ﺍﳌﻠﻔﹶﻖ‪ ،‬ﻭﳍﺬﺍ‬
‫ﻧﻔﻀﻞ ﰲ ﻫﺬﺍ ﺍﳌﻘﺎﻡ ﺍﺳﺘﺨﺪﺍﻡ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻷﻭﱃ ﻹﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫ﺇﻥ ﺍﻻﳓﺪﺍﺭ ﺍﳌﻠﻔﻖ ﻳﻌﱪ ﻋﻠﻰ ﺍﳊﺎﻟﺔ ﺍﻟﱵ ﻳﺘﻢ ﻓﻴﻬﺎ ﺇﺟﺮﺍﺀ ﻋﻤﻠﻴﺔ ﺗﻘﺪﻳﺮ‬
‫ﳌﺘﻐﲑﻳﻦ ﻻ ﻋﻼﻗﺔ ﳍﻤﺎ ﺑﺒﻌﻀﻬﻤﺎ ﺍﻟﺒﻌﺾ ﻣﻦ ﺍﻟﻨﺎﺣﻴﺔ ﺍﻟﻌﻤﻠﻴﺔ‪ .‬ﺗﻨﺘﺞ ﻋﻦ ﻫﺬﻩ‬
‫ﺍﻟﻮﺿﻌﻴﺔ ﻧﺘﺎﺋﺞ ﺇﺣﺼﺎﺋﻴﺔ ﺟﻴﺪﺓ ﻣﻦ ﺣﻴﺚ ﺍﻟﺪﻻﻟﺔ ﻋﺪﺍ ﺑﺮﻭﺯ ﻣﺸﻜﻞ ﺍﻻﺭﺗﺒﺎﻁ‬
‫ﺍﻟﺬﺍﰐ ﻟﻸﺧﻄﺎﺀ‪ .‬ﺇﻥ ﺗﻔﺴﲑ ﺍﳌﻌﻨﻮﻳﺔ ﺍﻟﻌﺎﻟﻴﺔ ﻟﻠﻨﺘﺎﺋﺞ ﻳﺮﺟﻊ ﳋﻀﻮﻉ ﺍﻟﺴﻠﺴﻠﺘﲔ ﻟﺘﺄﺛﲑ‬
‫ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﻟﻴﺲ ﻟﻌﻨﺼﺮ ﺍﻟﺴﺒﺒﻴﺔ‪.‬ﺳﻴﺘﻢ ﺗﻨﺎﻭﻝ ﻫﺬﺍ ﺍﳌﻮﺿﻮﻉ ﻣﻦ ﺯﺍﻭﻳﺔ‬
‫ﺃﻭﺳﻊ ﰲ ‪‬ﺎﻳﺔ ﺍﻟﻔﺼﻞ ﺍﳋﺎﻣﺲ ﻣﻦ ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ ﳍﺬﺍ ﺍﻟﻜﺘﺎﺏ ﲢﺖ ﻋﻨﻮﺍﻥ ﳕﺎﺫﺝ‬
‫ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﺍﻟﺸﻌﺎﻋﻴﺔ ﻭ ﺍﻻﺳﻘﺮﺍﺭﻳﺔ‪.‬‬

‫‪101‬‬
102
 ‫ﺍﻟﻔﺼﻞ ﺍﻟﺮﺍﺑﻊ‬

‫)‪(Seasonality‬‬ ‫ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ‬

‫ﺭﺃﻳﻨﺎ ﻓﻴﻤﺎ ﺳﺒﻖ‪ ،‬ﻛﻴﻔﻴﺔ ﺍﻟﺘﻌﺎﻣﻞ ﻣﻊ ﺳﻠﺴﻠﺔ ﺯﻣﻨﻴﺔ ﺫﺍﺕ ﻣﺮﻛﺒﺔ ﺍﲡﺎﻩ ﻋﺎﻡ‬
‫ﺇﺿﺎﻓﺔ ﺇﱃ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪ ،‬ﻟﻨﻨﺘﻘﻞ ﺍﻵﻥ ﺇﱃ ﻛﻴﻔﻴﺔ ﺍﻟﺘﻌﺎﻣﻞ ﰒ ﺍﻟﺘﻨﺒﺆ ﺑﻨﻤﻮﺫﺝ ﺳﻠﺴﻠﺔ‬
‫ﺯﻣﻨﻴﺔ ﺫﺍﺕ ﻣﺮﻛﺒﺔ ﻓﺼﻠﻴﺔ )ﺩﻭﺭﻳﺔ( ﺇﺿﺎﻓﺔ ﻟﻠﻤﺮﻛﺒﺘﲔ ﺍﻟﺴﺎﺑﻘﺘﲔ‪.‬‬
‫ﻋﺮ‪‬ﻓﻨﺎ ﺍﻟﻔﺼﻠﻴﺔ ﻫﻨﺎ ﺑﺎﻟﺘﺄﺛﲑﺍﺕ ﺍﳋﺎﺭﺟﻴﺔ ﻭﺍﳌﻨﺘﻈﻤﺔ ﺍﻟﱵ ﺗﻄﺮﺃ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﺗﱪﺯ ﻫﺬﻩ ﺍﳌﺮﻛﺒﺔ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ﻟﻸﺳﺒﺎﺏ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ -‬ﺍﻟﻌﻮﺍﻣﻞ ﺍﳌﺮﺗﺒﻄﺔ ﺑﺎﻟﻈﺮﻭﻑ ﺍﳉﻮﻳﺔ‪ ،‬ﻛﺎﻟﱪﻭﺩﺓ ﻭﺍﳊﺮﺍﺭﺓ ﺃﻭ ﺍﻟﺼﻴﻒ‬
‫ﻭﺍﻟﺸﺘﺎﺀ‪.‬‬
‫‪ -‬ﺍﻟﻌﻮﺍﻣﻞ ﺍﳌﺮﺗﺒﻄﺔ ﺑﺮ‪‬ﺯﻧﺎﻣﺔ ﺯﻣﻨﻴﺔ ﻛﺎﻷﻋﻴﺎﺩ ﻭﺍﳌﻨﺎﺳﺒﺎﺕ‪.‬‬
‫ﻛﻴﻔﻴﺔ ﺍﻟﺘﻌﺎﻣﻞ ﻣﻊ ﺍﻟﻔﺼﻠﻴﺔ ‪:‬‬
‫ﺭﺃﻳﻨﺎ ﰲ ﻫﺬﺍ ﺍﻟﺒﺎﺏ‪ ،‬ﺃﻧﻪ ﳝﻜﻦ ﺍﻟﺘﻌﺎﻣﻞ ﻣﻊ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺑﻄﺮﻳﻘﺘﲔ‪،‬‬
‫ﺍﻷﻭﱃ ﺗﺘﻤﺜﻞ ﰲ ﺇﺯﺍﻟﺔ ﻫﺬﻩ ﺍﳌﺮﻛﺒﺔ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ‪ ،‬ﰒﹼّ ﺍﻟﺘﻤﻬﻴﺪ ﻭﺍﻟﺘﻨﺒﺆ ﺑﺎﻟﺴﻠﺴﻠﺔ‬
‫ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﺍﻟﻨﺎﲡﺔ ﻓﻘﻂ‪ ،‬ﻭﻳﺘﻢ ﺍﻟﺘﻨﺒﺆ ﺍﻟﻨﻬﺎﺋﻲ ﰲ ﺍﻷﺧﲑ ﺑﺈﺿﺎﻓﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫‪.1‬‬
‫ﻛﻤﺎ ﳝﻜﻦ ﳕﺬﺟﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻣﺒﺎﺷﺮﺓ ﻭﻓﻖ ﳕﻮﺫﺝ ﻫﻮﻟﺖ ﺫﻭ ﺍﳌﻌﻠﻤﺘﲔ‬

‫‪ -1‬ﺟﺎﺀﺕ ﺍﻹﺯﺍﻟﺔ ﻫﻨﺎ ﰲ ﺳﻴﺎﻕﹴ ﺗﻘﲏ ﻟﺘﺴﻬﻴﻞ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻤﻬﻴﺪ ﺑﻴﻨﻤﺎ ﻗﺪ ﺗﺄﰐ ﻟﻜﺸﻒ ﺩﺭﺟﺔ‬
‫ﺍﻻﺭﺗﺒﺎﻁ ﺑﲔ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﰲ ﺣﺎﻟﺔ ﺍﻟﻨﻤﻮ‪ ،‬ﺣﻴﺚ ﺗﻈﻬﺮ ﻛﻠﻬﺎ ﻣﺘﺮﺍﺑﻄﺔ‬
‫)‪ (Autocorrelated‬ﺑﺴﺒﺐ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﺑﻴﻨﻤﺎ ﺇﺑﻌﺎﺩ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻳﺴﻤﺢ ﲟﻌﺮﻓﺔ‬
‫ﺃﺛﺮ ﺍﻟﺴﻴﺎﺳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﺍﳌﻄﺒﻘﺔ ﲟﻌﺰﻝ ﻋﻦ ﺍﻷﺛﺮ ﺍﻟﻔﺼﻠﻲ‪.‬‬
 ‫‪ -1‬ﺍﻹﺯﺍﻟﺔ ﻭ ﻃﺮﻗﻬﺎ‪:‬‬
‫ﰲ ﻧﻔﺲ ﺍﻟﺴﻴﺎﻕ ﻭﺑﻨﻔﺲ ﺍﳌﻨﻬﺠﻴﺔ ﺍﻟﺴﺎﺑﻘﺔ‪ ،‬ﻳﻜﻮﻥ ﺍﻟﺘﻌﺎﻣﻞ ﻣﻊ ﺍﳌﺮﻛﺒﺔ‬
‫ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﻓﻴﻤﻜﻦ ﺇﺯﺍﻟﺘﻬﺎ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﰒ ﺗﺮﺩ ﺇﻟﻴﻬﺎ ﰲ ﺍﻷﺧﲑ ﻟﻠﺤﺼﻮﻝ ﻋﻠﻰ ﺍﻟﺘﻮﻗﻊ‬
‫ﺍﻟﻨﻬﺎﺋﻲ ﺍﻟﺸﺎﻣﻞ ﻟﻜﻞ ﺍﳌﺮﻛﺒﺎﺕ ﺍﳌﻮﺟﻮﺩﺓ ﺃﺻﻼﹰ ﰲ ﺍﻟﺴﻠﺴﻠﺔ‪ .‬ﻛﻤﺎ ﳝﻜﻦ ﳕﺬﺟﺘﻬﺎ‬
‫ﻣﺒﺎﺷﺮﺓ ﻭﻓﻖ ﻃﺮﻳﻘﺔ "ﻫﻮﻟﺖ ﻭﻧﺘﺮﺯ" ﺫﺍﺕ ﺍﻟﺜﻼﺛﺔ ﻣﻌﺎﺩﻻﺕ ﻭﻣﻌﺎﻣﻼﺕ ﺃﻭ ﻃﺮﻳﻘﺔ‬
‫‪ Buys-Ballot‬ﰲ ﺣﺎﻻﺕ ﺧﺎﺻﺔ‪.‬‬
‫ﻓﻔﻲ ﺍﻟﺴﺎﺑﻖ‪ ،‬ﻭﻧﻈﺮ‪‬ﺍ ﻟﻌﺪﻡ ﺗﻄﻮﺭ ﺃﺟﻬﺰﺓ ﺍﳊﺴﺎﺏ ﺍﻹﻟﻜﺘﺮﻭﻧﻴﺔ‪ ،‬ﻛﺎﻥ ﻣﻦ‬
‫ﺍﻟﺼﻌﺐ ﲟﻜﺎﻥ ﳕﺬﺟﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻭﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺁﻧﻴ‪‬ﺎ‪ ،‬ﻛﻮ‪‬ﺎ ﺗﺘﻄﻠﺐ ﳎﻬﻮﺩﺍﺕ ﻣﻌﺘﱪﺓ‬
‫ﳊﺴﺎﺏ ﺍﳌﻌﺎﻣﻼﺕ ﺍﻟﺜﻼﺛﺔ ) ‪ ( b, g , a‬ﺍﻟﱵ ﺳﻨﺮﺍﻫﺎ ﻓﻴﻤﺎ ﺑﻌﺪ ﻭﺍﻟﱵ ﺗﺴﺘﻠﺰﻡ ﺗﻄﺒﻴﻖ‬
‫ﻃﺮﻳﻘﺔ ﺍﻟﺒﺤﺚ ﺍﻟﺸﺒﻜﻲ )‪ (Grid search‬ﻟﺘﺤﺪﻳﺪ ﻗﻴ‪‬ﻤﻬﺎ ﻭﺍﻟﱵ ﺗﻀﻤﻦ ﺗﺪﻧﻴﺔ‬
‫ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ )‪. (RSS‬‬
‫‪2‬‬

‫ﳍﺬﺍ ﺍﻟﺴﺒﺐ‪ ،‬ﻛﺎﻧﺖ ﻋﻤﻠﻴﺔ ﺍﻹﺯﺍﻟﺔ ﳌﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺫﺍﺕ ﺃﳘﻴﺔ‬
‫ﺑﺎﻟﻐﺔ ﰲ ﳎﺎﻝ ﺍﻟﻘﻴﺎﺱ ﺑﺸﻜﻞ ﻋﺎﻡ ﺃﻭ ﺍﻟﺘﻤﻬﻴﺪ ﻭﺍﻟﺘﻨﺒﺆ ﺑﺸﻜﻞ ﺧﺎﺹ‪.‬‬
‫ﻧﺴﺘﻌﺮﺽ ﻓﻴﻤﺎ ﻳﻠﻲ ﳎﻤﻮﻋﺔ ﻣﻦ ﺍﻟﻄﺮﻕ ﺍﳌﺴﺘﻌﻤﻠﺔ ﳍﺬﺍ ﺍﻟﻐﺮﺽ‪ ،‬ﻭﺍﻟﱵ ﳝﻜﻦ‬
‫ﺗﻘﺴﻴﻤﻬﺎ ﺇﱃ ﻓﺌﺘﲔ ﺭﺋﻴﺴﺘﲔ‪ ،‬ﺍﻷﻭﱃ ﻻ ﲢﺴﺐ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ ﻭﺍﻟﺜﺎﻧﻴﺔ ﲢﺴﺒﻬﺎ‪.‬‬

‫ﺍﻟﻄﺮﻕ ﺍﻟﱵ ﻻ ﲢﺴﺐ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ‪:‬‬

‫‪‬ﺘﻢ ﻫﺬﻩ ﺍﻟﺘﻘﻨﻴﺎﺕ ﺑﺈﺯﺍﻟﺔ ﺍﳌﺮﻛﺒﺔ ﺍﻟﺪﻭﺭﻳﺔ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﳌﺪﺭﻭﺳﺔ‬
‫ﺩﻭﻥ ﺇﻋﻄﺎﺀ ﺃﻱ ﺃﳘﻴﺔ ﻟﻌﻤﻠﻴﺔ ﺗﻘﺪﻳﺮ ﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﺣﻴﺚ ﻳﺘﻢ ﺍﳊﺼﻮﻝ ﰲ‬
‫‪‬ﺎﻳﺔ ﺍﳌﻄﺎﻑ ﻋﻠﻰ ﺳﻠﺴﻠﺔ ﺧﺎﻟﻴﺔ ﻣﻦ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ‪.‬‬

‫‪ - 2‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ‪ :‬ﻣﻘﺪﻣﺔ ﰲ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ‪ ،‬ﻉ‪.‬ﻉ‪ .‬ﺷﺮﻳﻒ‪ ،‬ﺹ‪.33 ،32‬‬

‫‪104‬‬
 ‫‪ -1‬ﻃﺮﻳﻘﱵ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﻟﺒﺴﻴﻄﺔ ﻭﺍﳌﻤﺮﻛﺰﺓ‪:‬‬
‫ﻫﺎﺗﺎﻥ ﺍﻟﻄﺮﻳﻘﺘﺎﻥ‪ ،‬ﻭﺍﻟﻠﺘﺎﻥ ﺭﺃﻳﻨﺎﳘﺎ ﻣﻦ ﻗﺒﻞ ﺻﺎﳊﺘﺎﻥ ﻛﺬﻟﻚ ﻹﺯﺍﻟﺔ ﺍﻟﻔﺼﻠﻴﺔ‬
‫ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﻟﺘﻮﺿﻴﺢ ﺫﻟﻚ ﻧﻔﺘﺮﺽ ﺃﻧﻪ ﻟﺪﻳﻨﺎ ﺳﻠﺴﻠﺔ ﺯﻣﻨﻴﺔ‬
‫ﺫﺍﺕ ﻣﺮﻛﺒﺘﲔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻭﰲ ﺷﻜﻞ ﲡﻤﻴﻌﻲ ﺃﻱ‪:‬‬
‫‪yt = Lt + et‬‬
‫‪E( e t ) 2 = s 2‬‬ ‫ﻭ‬ ‫‪E( e t ) = 0‬‬ ‫ﺑـ‬
‫ﻭ ‪var( y t ) = s 2‬‬ ‫‪E( y t ) = L t‬‬ ‫ﻣﻨﻪ‬
‫~‬
‫ﻓﺈﺫﺍ ﺍﺳﺘﻌﻨﺎ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﻟﺒﺴﻴﻄﺔ ﻓﺈﻥ ﺗﺒﺎﻳﻦ ‪ y t‬ﻳﻜﻮﻥ‪:‬‬
‫~‬ ‫‪1 n -1‬‬
‫(‪E( y t )2 = E‬‬ ‫‪å y t-r ) 2‬‬
‫‪n r=0‬‬
‫‪1‬‬ ‫‪s2‬‬
‫=‬ ‫‪2‬‬
‫= ) ‪( ns 2‬‬
‫‪n‬‬ ‫‪n‬‬

‫ﻣﻨﻪ ﻓﺘﺒﺎﻳﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﻤﻬﺪﺓ ﻳﻜﻮﻥ ﺃﻗﻞ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ‪،‬‬

‫ﺇﻻ ﺃﻥ ﺍﻟﻄﺮﻳﻘﺘﺎﻥ ﻻ ‪‬ﺘﻤﺎﻥ ﲝﺴﺎﺏ ﺍﳌﺆﺷﺮﺍﺕ ﺃﻱ ﺍﳌﻌﺎﻣﻼﺕ ﺍﻟﻔﺼﻠﻴﺔ‬
‫)‪.(Seasonal indices‬‬

‫‪ -2‬ﻃﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ‪:‬‬
‫ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺗﺼﻠﺢ ﻛﺬﻟﻚ ﻹﺯﺍﻟﺔ ﺍﻟﺪﻭﺭﻳﺔ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﺗﻜﺘﺐ‬
‫ﺭﻳﺎﺿﻴ‪‬ﺎ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪Dp y t = y y - y t - p‬‬ ‫)‪(4.1‬‬

‫‪= (1 - L p ) y t‬‬
‫ﻋﻠﻰ‬ ‫‪ p = 4‬ﻭ ‪p = 12‬‬ ‫ﺣﻴﺚ ﺗﻜﻮﻥ ﰲ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﻔﺼﻠﻴﺔ ﻭﺍﻟﺸﻬﺮﻳﺔ‬
‫ﻛﻤﺎ‬ ‫ﺍﻟﺘﺮﺗﻴﺐ‪ .‬ﻟﻘﺪ ﺍﺳﺘﺨﺪﻣﺖ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻹﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﳌﺎ ‪p = 1‬‬
‫ﺭﺃﻳﻨﺎ ﺳﺎﺑﻘﺎﹰ‪.‬‬

‫‪105‬‬
‫ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ )ﺇﺯﺍﻟﺔ ﺍﻟﻔﺼﻠﻴﺔ( ﺗﺘﻢ ﺑﻄﺮﻳﻘﺔ ﻋﻜﺴﻴﺔ ﻭﳝﻜﻦ‬
‫ﺗﻠﺨﻴﺼﻬﺎ ﰲ ﺍﳌﺮﺍﺣﻞ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫ﺃ‪ -‬ﻳﺘﻢ ﰲ ﻋﻼﻗﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ) ‪ ( p‬ﺍﺳﺘﺒﺪﺍﻝ ) ‪ ( t‬ﺏ )‪(T+1‬‬
‫ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪D‬‬ ‫‪p‬‬
‫‪y t = y ta = y t - y t -‬‬ ‫‪p‬‬

‫ﺑﻌﺪ ﺍﻟﺘﻌﻮﻳﺾ ﺗﺼﺒﺢ‬

‫‪yˆ Ta +1 = yˆ T +1 - y T - p +1‬‬

‫ﺑـ‪ -‬ﺍﻟﺘﻨﺒﺆ ﺑﻄﺮﻳﻘﺔ ﻣﻼﺋﻤﺔ ﻟﻠﺤﺼﻮﻝ ﻋﻠﻰ ) ‪ ، ( yˆ Ta+1‬ﻛﺄﻥ ﺗﻜﻮﻥ ﺑﺎﺳﺘﻌﻤﺎﻝ‬

‫ﳕﻮﺫﺝ ﺍﻟﺘﻤﻬﻴﺪ ﺍﻵﺳﻲ ﺍﻷﺣﺎﺩﻱ ﺇﻥ ﺑﻘﻴﺖ ﺍﻟﺴﻠﺴﻠﺔ ﺑﻌﺪ ﺍﻹﺯﺍﻟﺔ ﻋﺸﻮﺍﺋـﻴﺔ ﻓﻘﻂ‪.‬‬
‫ﺟـ‪ -‬ﺣﺴﺎﺏ ﺍﻟﺘﻨﺒﺆ ‪ ، yˆT+1‬ﺇﺫﹰﺍ‪:‬‬
‫‪yˆ T +1 = yˆ Ta +1 + yT - p +1‬‬

‫ﻟﺬﺍ ﻳﺒﻘﻰ ﺗﻨﺒﺆ ‪ yˆT+l‬ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ﻣﺮﻫﻮﻥ ﲟﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬

‫ﻭﻋﻼﻗﺎ‪‬ﺎ ﺑﺒﻌﻀﻬﺎ‪.‬‬
‫)‪(22‬‬ ‫ﻣﺜﺎﻝ ‪ :20‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﻧﻔﺲ ﻣﻌﻄﻴﺎﺕ ﺍﻻﺳﺘﻬﻼﻙ ﺍﻟﻔﺼﻠﻴﺔ ﺍﳌﻤﺜﻠﺔ ﰲ ﺍﳉﺪﻭﻝ‬
‫ﺍﻟﺘﺎﱄ‪ ،‬ﺍﳌﻄﻠﻮﺏ ﺇﺯﺍﻟﺔ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺍﻟﱵ ﺗﻈﻬﺮ ﰲ ﺷﻜﻞ ﻗﻤﻢ ﻣﻨﺘﻈﻤﺔ ﺑﺪﻭﺭﻳﺔ‬
‫ﻣﻘﺪﺍﺭﻫﺎ )‪ (4‬ﻓﺼﻮﻝ ﺑﻄﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ‪.‬‬
‫ﺍﻟﺸﻜﻞ )‪:(20‬ﺳﻠﺴﻠﺔ ﺫﺍﺕ ﻣﺮﻛﺒﺔ ﻓﺼﻠﻴﺔ‬

‫‪106‬‬
‫ﻛﻮﻥ ﺍﳌﻌﻄﻴﺎﺕ ﻓﺼﻠﻴﺔ ﳔﺘﺎﺭ ‪ p = 4‬ﻟﺘﻄﺒﻴﻖ ﺗﻘﻨﻴﺔ ﺍﻟﻔﺮﻭﻕ ﺍﳌﻨﺎﺳﺒﺔ ﻛﻤﺎ ﻳﺘﺒﲔ‬
‫ﺿﻤﻦ ﺍﻟﻌﻤﻠﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ ﺍﻟﱵ ﻳﻌﻜﺴﻬﺎ ﺍﳉﺪﻭﻝ )‪ (21‬ﺍﳌﻮﺍﱄ‪:‬‬

‫اﻟﺠﺪول )‪ :(21‬ﺟﺪﻭﻝ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻟﺮﺍﺑﻌﺔ‬

‫‪C ta = c t - c t - 4‬‬ ‫‪c t -4‬‬ ‫‪c‬‬ ‫‪t‬‬ ‫‪T‬‬
‫‪-‬‬ ‫‪-‬‬ ‫‪155‬‬ ‫‪1‬‬
‫‪-‬‬ ‫‪-‬‬ ‫‪158‬‬ ‫‪2‬‬
‫‪-‬‬ ‫‪-‬‬ ‫‪168‬‬ ‫‪3‬‬
‫‪-‬‬ ‫‪-‬‬ ‫‪171‬‬ ‫‪4‬‬
‫‪2-‬‬ ‫‪155‬‬ ‫‪153‬‬ ‫‪5‬‬
‫‪2-‬‬ ‫‪158‬‬ ‫‪156‬‬ ‫‪6‬‬
‫‪1-‬‬ ‫‪163‬‬ ‫‪162‬‬ ‫‪7‬‬
‫‪1‬‬ ‫‪171‬‬ ‫‪172‬‬ ‫‪8‬‬
‫‪9‬‬ ‫‪153‬‬ ‫‪162‬‬ ‫‪9‬‬
‫‪8‬‬ ‫‪156‬‬ ‫‪164‬‬ ‫‪10‬‬
‫‪11‬‬ ‫‪162‬‬ ‫‪173‬‬ ‫‪11‬‬
‫‪9‬‬ ‫‪172‬‬ ‫‪181‬‬ ‫‪12‬‬

‫? = ‪t = 1 : c 1 = 155 , c t - 4 = c -3‬‬ ‫ﺣﻴﺚ ﳌﺎ‪:‬‬

‫? = ‪t = 4 : c 4 = 171 , c t - 4 = c 0‬‬ ‫ﺇﱃ ﻏﺎﻳﺔ‬
‫‪t = 5: c 5 = 153 , c t - 4 = c 1 = 155‬‬ ‫ﺑﻴﻨﻤﺎ ﳌﺎ‬
‫ﲟﻌﲎ ﺃﻧﻨﺎ ﻧﻀﻴ‪‬ﻊ ) ‪ ( p‬ﻣﻦ ﺍﳌﺸﺎﻫﺪﺍﺕ ﻋﻨﺪ ﺗﻄﺒﻴﻖ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺃﻱ‪:‬‬
‫‪C ta = c t - c t - 4‬‬

‫ﺧﺎﻟﻴﺔ ﻣﻦ ﺍﻟﻔﺼﻠﻴﺔ ﻛﻤﺎ ﻳﻮﺿﺤﻪ ﺍﻟﺒﻴﺎﻥ ﺍﻟﺘﺎﱄ‪:‬‬ ‫‪C ta‬‬ ‫ﺗﻜﻮﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻨﺎﲡﺔ‬

‫‪107‬‬
 ‫ﺍﻟﺸﻜﻞ )‪ :(21‬ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﻨﺰﻭﻋﺔ ﺍﻟﻔﺼﻠﻴﺔ‬

‫ﺑﺪﺍﻟﺔ ﺍﲡﺎﻩ ﻋﺎﻡ ﺧﻄﻲ ﻛﺎﻵﰐ‪:‬‬ ‫‪C ta‬‬ ‫ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻦ ﻫﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ‬
‫‪C ta = a + b t + u t‬‬

‫ﺑﺘﻘﺪﻳﺮ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻟﺔ ﺑﺎﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﳓﺼﻞ ﻋﻠﻰ‪:‬‬

‫‪Cˆ ta = -13.78 + 2.1t‬‬
‫)‪(3.25) (0.37‬‬
‫ﻣﺎ ﺑﲔ ﺍﻟﻘﻮﺳﲔ ﳝﺜﻞ ﺍﻻﳓﺮﺍﻓﺎﺕ ﺍﳌﻌﻴﺎﺭﻳﺔ ﻟﻠﻤﻌﻠﻤﺔ‪ ،‬ﻭﺗﻨﺒﺆ ﺍﻟﻔﺘﺮﺓ ﺍﻟﻘﺎﺩﻣﺔ‬
‫ﻫﻮ‪:‬‬
‫)‪Cˆ Ta +1 = - 13 . 78 + 2.1(T + 1‬‬

‫‪Cˆ 13a = -13.78 + 2.1(13) = 13.52‬‬

‫ﻫﺬﻩ ﺍﻟﻨﺘﻴﺠﺔ ﺗﻌﺘﱪ ﻛﺘﻨﺒﺆ ﲟﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻓﻘﻂ‪ ،‬ﰲ ﺣﲔ ﳛﺘﺎﺝ ﺍﻟﺪﺍﺭﺱ‬
‫ﻟﺘﻨﺒﺆ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ ﺃﻱ‪:‬‬
‫‪Cˆ Ta+1 = cˆT +1 - cT -3‬‬

‫‪cˆT +1 = Cˆ Ta+1 - cT -3‬‬

‫‪cˆ13 = Cˆ 13a - c 9‬‬

‫‪cˆ13 = 13.52 + 162 = 175.52‬‬

‫‪108‬‬
 ‫ﺍﻟﻄﺮﻕ ﺍﻟﱵ ﺗﺰﻳﻞ ﺍﻟﻔﺼﻠﻴﺔ ﻣﻊ ﺣﺴﺎﺏ ﻣﺆﺷﺮﺍ‪‬ﺎ‪:‬‬
‫ﻧﺴﺘﻌﲔ ﰲ ﻫﺬﺍ ﺍ‪‬ﺎﻝ ﺑﺎﻟﻄﺮﻕ ﺍﻟﻌﺎﻣﺔ ﺍﻟﱵ ﲢﺴﺐ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ ﰒ‬
‫ﺗﺰﻳﻞ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻟﻠﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺑﻌﻼﻗﺔ ﺟﺪﺍﺋﻴﺔ ﺃﻭ ﲡﻤﻴﻌﻴﺔ ﳌﺮﻛﺒﺎ‪‬ﺎ‪.‬‬
‫‪ -1‬ﻃﺮﻳﻘﺔ ﺍﻟﻨﺴﺐ ﺍﳌﻮﲰﻴﺔ‪:‬‬
‫ﺗﺴﺘﻌﻤﻞ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺍﳉﺪﻭﻝ ﻭﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﺍﻟﻌﺎﻡ ﳊﺴﺎﺏ ﺍﳌﺆﺷﺮﺍﺕ‬
‫ﺍﻟﻔﺼﻠﻴﺔ )ﺍﳌﻮﲰﻴﺔ( ﺑﺄﺳﻠﻮﺏ ﻻ ﻳﻔﺮﻕ ﺑﲔ ﺍﻟﺸﻜﻞ ﺍﳉﺪﺍﺋﻲ ﻭﺍﻟﺘﺠﻤﻴﻌﻲ ﺃﺛﻨﺎﺀ‬
‫ﺍﳊﺴﺎﺏ‪ ،‬ﻭﳝﻜﻦ ﺗﻠﺨﻴﺺ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻓﻴﻤﺎ ﻳﻠﻲ‪:‬‬
‫ﺑﻌﺪ ﺗﺼﻨﻴﻒ ﺍﳌﺸﺎﻫﺪﺍﺕ ﰲ ﺷﻜﻞ ﺟﺪﻭﱄ ﺃﻳﻦ ﲤﺜﻞ ﺍﻷﺳﻄﺮ ﺍﻟﺴﻨﻮﺍﺕ‬
‫ﻭﺍﻷﻋﻤﺪﺓ ﺍﻟﺪﻭﺭﻳﺔ )ﻓﺼﻞ‪ ،‬ﺷﻬﺮ‪...‬ﺍﱁ(‪ ،‬ﻳﺘﻢ ﻣﺎ ﻳﻠﻲ‪:‬‬
‫‪ -‬ﺣﺴﺎﺏ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﻟﻠﻤﺸﺎﻫﺪﺍﺕ ﺍﳌﻘﺎﺑﻠﺔ ﻟﻜﻞ ﺳﻨﺔ ‪. i‬‬
‫‪p‬‬
‫_‬
‫‪1‬‬ ‫_‬
‫= ‪yi‬‬
‫‪p‬‬
‫‪å y ij‬‬
‫‪j=1‬‬
‫)‪(4.2‬‬

‫‪i = 1, 2,..., m‬‬

‫‪j = 1, 2,..., p‬‬
‫ﺣﻴﺚ ‪ i‬ﺩﻟﻴﻞ ﺍﻟﺴﻨﺔ ﻭ ‪ j‬ﺩﻟﻴﻞ ﺍﳌﻮﲰﻴﺔ ﻣﺜﻼ ‪.‬‬
‫‪ -‬ﺣﺴﺎﺏ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﺍﻟﻌﺎﻡ‬
‫‪m‬‬
‫=‬ ‫‪1‬‬ ‫_‬
‫=‪y‬‬
‫‪m‬‬
‫‪åy‬‬
‫‪i=1‬‬
‫‪i‬‬
‫)‪(4.3‬‬

‫ﻛﻤﺎ ﳝﻜﻦ ﻛﺘﺎﺑﺘﻪ ﻭﻓﻖ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‬

‫‪1 p‬‬
‫= ‪y‬‬ ‫‪å yj‬‬
‫‪p j=1‬‬

‫ﺣﺴﺎﺏ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﻟﻜﻞ ﻓﺼﻞ ﺃﻭ ﺷﻬﺮ ﻋﱪ ﺍﻟﺴﻨﻮﺍﺕ‪ ،‬ﻛﺄﻥ ﳒﻤﻊ‬

‫ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﳋﺎﺻﺔ ﺑﺎﻟﻔﺼﻞ )ﺍﻟﺸﻬﺮ( ‪ j‬ﻟﻜﻞ ﺍﻟﺴﻨﻮﺍﺕ ﻭﻧﻘﺴﻤﻪ ﻋﻠﻰ ﻋﺪﺩ ﻫﺬﻩ‬
‫ﺍﻷﺧﲑﺓ ‪ m‬ﺃﻱ‪:‬‬

‫‪109‬‬
 ‫‪m‬‬
‫_‬
‫‪1‬‬ ‫_‬
‫= ‪yj‬‬
‫‪m‬‬
‫‪å y ij‬‬
‫‪i=1‬‬
‫)‪(4.4‬‬

‫‪ -‬ﺣﺴﺎﺏ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ‪j‬‬

‫_‬
‫‪yj‬‬
‫=‪Sj‬‬ ‫»‬
‫)‪(4.5‬‬
‫‪y‬‬
‫‪ -‬ﺣﺴﺎﺏ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﻟﻔﺼﻠﻴﺔ ﻭﻛﻤﺎ ﻳﻠﻲ‬
‫‪y ij‬‬
‫= ‪ya‬‬ ‫)‪(4.6‬‬
‫‪ij S j‬‬

‫ﻭﺍﻟﺘﻨﺒﺆ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻳﺘﻢ ‪‬ﺬﺍ ﺍﻟﺸﻜﻞ‬

‫‪y T +1, j‬‬
‫= ‪y aT +1, j‬‬ ‫)‪(4.7‬‬
‫‪Sj‬‬

‫ﺃﻳﻦ ﺃﹸﺿﻴﻒ ‪ j‬ﻫﻨﺎ ﻟﻐﺮﺽ ﺍﺳﺘﺒﻴﺎﻥ ﺍﳌﺸﺎﻫﺪﺓ ﺍﳌﻘﺎﺑﻠﺔ ﻟﻠﻤﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ‬

‫ﺍﳌﻌﲏ ‪ j‬ﺃﻱ ﺭﺑﻂ ﺍﳌﺸﺎﻫﺪﺓ ﺑﺎﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ﺍﳌﻘﺎﺑﻞ ﳍﺎ‪ ،‬ﻭﻣﻨﻪ‬
‫‪yˆ T +1, j = yˆ a T +1, j . S j‬‬ ‫)‪(4.8‬‬

‫ﻭﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫‪yˆT + L, j = yˆ a T + L, j . S j‬‬ ‫)‪(4.9‬‬

‫ﻣﺜﺎﻝ ‪ :21‬ﳝﻜﻦ ﺗﻮﺿﻴﺢ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺑﺎﺳﺘﻌﻤﺎﻝ ﻧﻔﺲ ﻣﻌﻄﻴﺎﺕ ﺍﻻﺳﺘﻬﻼﻙ‬

‫ﻭﺍﳌﻤﺜﻠﺔ ﰲ ﺷﻜﻞ ﺟﺪﻭﱄ ﻭﻛﻤﺎ ﻳﺄﰐ‪:‬‬
‫ﺍﳉﺪﻭﻝ)‪ (22‬ﻣﻌﻄﻴﺎﺕ ﺍﻻﺳﺘﻬﻼﻙ ﺍﻟﻔﺼﻠﻴﺔ‬
‫‪-‬‬
‫‪yi‬‬ ‫‪.4‬‬ ‫‪.3‬‬ ‫‪.2‬‬ ‫‪.1‬‬ ‫ﺍﻟﺴﻨﺔ‪/‬ﺍﻟﻔﺼﻞ‬
‫‪161.75‬‬ ‫‪171‬‬ ‫‪163‬‬ ‫‪158‬‬ ‫‪155‬‬ ‫‪1‬‬
‫‪160.75‬‬ ‫‪172‬‬ ‫‪162‬‬ ‫‪156‬‬ ‫‪153‬‬ ‫‪2‬‬
‫‪170‬‬ ‫‪181‬‬ ‫‪173‬‬ ‫‪164‬‬ ‫‪162‬‬ ‫‪3‬‬
‫‪-‬‬
‫‪164.16‬‬ ‫‪174.66‬‬ ‫‪166‬‬ ‫‪159.33‬‬ ‫‪156.66‬‬ ‫‪yi‬‬
‫‪1.06‬‬ ‫‪1.01‬‬ ‫‪0.97‬‬ ‫‪0.95‬‬ ‫‪Sj‬‬

‫‪110‬‬
 ‫=‬

‫ﺣﻴﺚ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﺍﻟﻌﺎﻡ ﻣﻘﺪﺭ ﺏ ‪y = 164 .16‬‬

‫ﲢﺴﺐ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻘﺴﻤﺔ ﻛﻞ ﻣﺸﺎﻫﺪﺓ ﻋﻠﻰ ﺍﳌﺆﺷﺮ‬
‫ﺍﻟﻔﺼﻠﻲ ﺍﳌﻘﺎﺑﻞ‪:‬‬
‫‪y ij‬‬
‫= ‪y aij‬‬ ‫)‪(4.10‬‬
‫‪Sj‬‬

‫ﺍﳉﺪﻭﻝ )‪ :(23‬ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﻟﻔﺼﻠﻴﺔ‬

‫‪0.4‬‬ ‫‪0.3‬‬ ‫‪0.2‬‬ ‫‪0.1‬‬ ‫ﺍﻟﺴﻨﺔ‪/‬ﺍﻟﻔﺼﻞ‬
‫‪161.32‬‬ ‫‪161.39‬‬ ‫‪162.88‬‬ ‫‪163.16‬‬ ‫‪1‬‬
‫‪162.26‬‬ ‫‪160.04‬‬ ‫‪160.82‬‬ ‫‪161.05‬‬ ‫‪2‬‬
‫‪170.75‬‬ ‫‪171.29‬‬ ‫‪169.07‬‬ ‫‪170.52‬‬ ‫‪3‬‬

‫ﻟﻐﺮﺽ ﺍﻟﺘﻨﺒﺆ ﺍﳌﺴﺘﻘﺒﻠﻲ ﻧﺴﺘﻄﻴﻊ ﻛﻤﺎ ﰲ ﺍﻟﺴﺎﺑﻖ ﻛﺘﺎﺑﺔ ﺍﻟﻌﻼﻗﺔ ﰲ ﺷﻜﻞ‬

‫ﺟﺪﻳﺪ ﺑﺪﻻﻟﺔ ‪ t‬ﻭ ‪ j‬ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪yt , j‬‬
‫= ‪ya t, j‬‬ ‫)‪(4.11‬‬
‫‪Sj‬‬
‫ﺣﻴﺚ ﻧﺴﺘﻄﻴﻊ ﺍﻟﺮﺑﻂ ﺑﲔ ‪ i j , t‬ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫‪t = ( i -1 )p+ j‬‬ ‫)‪(4.12‬‬

‫ﻭﺍﻟﺘﻨﺒﺆ ﺣﺴﺐ‬ ‫‪t = 13‬‬ ‫ﻭ ‪ j = 1‬ﻓﺈﻥ‬ ‫‪i=4‬‬ ‫ﰲ ﺣﺎﻟﺔ ﺍﳉﺪﻭﻝ ﺍﻟﺴﺎﺑﻖ‪ ،‬ﻭﳌﺎ‬
‫ﻫﺬﺍ ﺍﳌﺜﺎﻝ ﳛﺴﺐ ﺑﺎﻟﻄﺮﻳﻘﺔ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫‪yˆ T +1,1 = yˆ a T +1,1 . S1‬‬ ‫)‪(4.13‬‬

‫ﺣﻴﺚ ﳚﺐ ﺍﻟﺘﻨﺒﺆ ﺑﺎﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﻟﻔﺼﻠﻴﺔ ﻟﻠﻔﺘﺮﺓ ﺍﻟﻘﺎﺩﻣﺔ ﻭﺑﻄﺮﻳﻘﺔ‬

‫ﻣﻼﺋﻤﺔ ﺑﻌﺪ ﲢﻠﻴﻞ ﺍﳌﺮﻛﺒﺎﺕ ﺍﳌﺘﺒﻘﻴﺔ ﻓﻴﻬﺎ‪ ،‬ﺃﻳﻦ )‪ (T+1‬ﺗﻮﺍﻓﻖ ﻫﻨﺎ ﺍﻟﻔﺘﺮﺓ "‪ "13‬ﻭﻫﻲ‬
‫ﺗﻨﺒﺆ ﺍﻟﻔﺼﻞ ﺍﻷﻭﻝ ﻟﻠﺴﻨﺔ ﺍﻟﺮﺍﺑﻌﺔ ﻭﻫﻜﺬﺍ‪.‬‬

‫‪111‬‬
 ‫‪ -2‬ﻃﺮﻳﻘﺔ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﻟﻨﺴﺒﻴﺔ‪:‬‬
‫ﺭﻏﻢ ﻛﺜﺮﺓ ﻣﺮﺍﺣﻞ ﺣﺴﺎ‪‬ﺎ‪ ،‬ﺇﻻ ﺃ‪‬ﺎ ﲤﺘﺎﺯ ﻋﻦ ﺳﺎﺑﻘﺘﻬﺎ ﰲ ﺃ‪‬ﺎ ﺗﺴﺘﻄﻴﻊ‬
‫ﺍﻟﺘﻔﺮﻗﺔ ﺑﲔ ﺍﻟﺸﻜﻞ ﺍﳉﺪﺍﺋﻲ ﻭﺍﻟﺘﺠﻤﻴﻌﻲ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬
‫ﻟﻠﺘﺬﻛﲑ‪ ،‬ﻧﻘﻮﻝ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺗﺘﻜﻮﻥ ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ﻣﻦ ﺃﺭﺑﻌﺔ‬
‫ﻣﺮﻛﺒﺎﺕ‪ ،‬ﻭﺍﻟﱵ ﻧﺼﻐﻬﺎ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬

‫) ‪y ij = f ( L ij , C ij , S ij , I ij‬‬ ‫)‪(4.14‬‬

‫ﲢﺖ ﻓﺮﺿﻴﺔ ﺛﺒﺎﺕ ﺍﻟﺘﺄﺛﲑ ﺍﻟﻔﺼﻠﻲ )ﲟﻌﲎ ﺃﻥ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ﺍﻷﻭﻝ ﳍﺬﻩ‬
‫ﺍﻟﺴﻨﺔ ﻳﺴﺎﻭﻱ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ﺍﻷﻭﻝ ﻟﻠﺴﻨﺔ ﺍﳌﺎﺿﻴﺔ ﺃﻭ ﺍﻟﻘﺎﺩﻣﺔ( ﰲ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ‪،‬‬
‫) ‪y ij = f ( L ij , C ij , S j , I ij‬‬ ‫ﺇﺫﺍﹰ‪:‬‬

‫ﺣﻴﺚ ﳝﻜﻦ ﺍﻵﻥ ﺗﻠﺨﻴﺺ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﰲ ﺍﳌﺮﺍﺣﻞ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬

‫‪3‬‬

‫‪ -1‬ﺍﳊﺎﻟﺔ ﺍﳉﺪﺍﺋﻴﺔ‪:‬‬
‫ﺑﻌﺪ ﲢﺪﻳﺪ ﺍﻟﻄﺒﻴﻌﺔ ﺍﳉﺪﺍﺋﻴﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻧﻜﺘﺐ ﻣﻌﺎﺩﻟﺘﻬﺎ ﻛﻤﺎ ﻳﺄﰐ‪:‬‬
‫‪y ij = L ij . C ij . S j . I ij‬‬ ‫)‪(4.15‬‬

‫‪ - 1-1‬ﺑﺘﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﳌﻤﺮﻛﺰﺓ )ﻡ‪.‬ﻡ‪.‬ﻡ( ﻟﻠﺘﻤﻬﻴﺪ ﺍﻟﱵ‬

‫ﺗﺰﻳﻞ ﺍﳊﻮﺍﺩﺙ ﺍﻟﻌﺎﺭﺿﺔ ﻣﻨﻬﺎ‪ ،‬ﺗﻌﺰﻝ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ‪،‬‬
‫ﻭﺗﻜﺘﺐ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻨﺎﲡﺔ ﻫﻜﺬﺍ‪:‬‬
‫~‬
‫‪y ij = L ij . C ij‬‬ ‫)‪(4.16‬‬

‫‪ - 3‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ ‪:‬‬

‫‪Econometric Models and Economic Forecast, R..Pindyck , D. Rubinfeld, p.487‬‬

‫‪112‬‬
 ‫‪ - 2-1‬ﺗﻘﺴﻴﻢ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ ﻋﻠﻰ ﺍﻟﻨﺎﲡﺔ ﺃﻱ‬
‫‪y ij‬‬ ‫‪L ij . C ij . S j . I ij‬‬
‫= ‪z ij‬‬ ‫=‬ ‫)‪(4.17‬‬
‫~‬
‫‪y ij‬‬ ‫‪L ij . C ij‬‬

‫ﲝﻴﺚ ﻻ ﲢﻮﻱ ﺍﻟﻨﺎﲡﺔ ﺳﻮﻯ ﻋﻠﻰ ﺍﳌﺮﻛﺒﺘﲔ ﺍﻟﻔﺼﻠﻴﺔ ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻓﻘﻂ ﺃﻱ‪:‬‬
‫‪z ij = S j. I ij‬‬

‫‪ -3-1‬ﺣﺴﺎﺏ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ‪ ،‬ﻭﻳﺘﻢ ﺑﻘﺴﻤﺔ ﻣﺸﺎﻫﺪﺍﺕ ﺍﻟﻔﺼﻞ" ‪" j‬‬

‫ﺍﳋﺎﺻﺔ ﺑﻜﻞ ﺳﻨﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻷﺧﲑﺓ ﻋﻠﻰ ﻋﺪﺩ ﺍﻟﺴﻨﻮﺍﺕ ﻧﺎﻗﺺ ﻭﺍﺣﺪ ) ‪( m - 1‬‬
‫ﻳﺮﺟﻊ ﻫﺬﺍ ﺍﻟﻔﻌﻞ ﺇﱃ ﺳﺒﺐ ﻓﻘﺪﺍﻥ ﻣﻦ ﻛﻞ ﻋﻤﻮﺩ ﻣﺸﺎﻫﺪﺓ ﺑﻔﻌﻞ ﺗﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ‬
‫ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﳌﻤﺮﻛﺰﺓ‪.‬‬
‫‪1 m‬‬
‫= ‪Sj‬‬ ‫‪å zij‬‬
‫‪m - 1 i =1‬‬
‫)‪(4.18‬‬

‫‪j = 1, 2,... , p‬‬

‫‪ - 4-1‬ﻳﺸﺘﺮﻁ ﺃﻥ ﻳﻌﺎﺩﻝ ﳎﻤﻮﻉ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ ﻋﺪﺩ ﻣﺸﺎﻫﺪﺍﺕ ﻛﻞ‬
‫ﺳﻨﺔ ﺃﻱ ) ‪.( p‬‬
‫‪p‬‬

‫‪åS‬‬
‫‪j =1‬‬
‫‪j‬‬ ‫‪=p‬‬ ‫)‪(4.19‬‬

‫ﰲ ﺣﺎﻟﺔ ﻋﺪﻡ ﲢﻘﻖ ﺍﻟﺸﺮﻁ ﳒﺮﻱ ﻋﻤﻠﻴﺔ ﺍﻟﺘﺤﻮﻳﻞ ﺍﻟﺘﺎﻟﻴﺔ ﳊﺴﺎﺏ ﺍﳌﺆﺷﺮﺍﺕ‬
‫ﺍﻟﻔﺼﻠﻴﺔ ﺍﳌﻌﺪﻟﺔ‪.‬‬
‫ﻓﺈﺫﺍ ﻛﺎﻥ‬
‫‪p‬‬
‫‪åS j = x ¹ p‬‬
‫‪j= 1‬‬
‫‪p‬‬
‫) (‪s j = S j‬‬ ‫ﻓﺈﻥ‬
‫‪x‬‬
‫‪Sj‬‬
‫= ‪sj‬‬ ‫_‬ ‫ﺃﻳﻦ‬
‫‪S‬‬

‫‪113‬‬
 ‫_‬
‫=‪S‬‬
‫‪x‬‬
‫=‬
‫‪åS‬‬ ‫‪j‬‬
‫ﻭ‬
‫‪p‬‬ ‫‪p‬‬

‫‪ -5-1‬ﻳﺘﻢ ﺣﺴﺎﺏ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﳌﻮﲰﻴﺔ ﺑﻘﺴﻤﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ‬

‫ﻋﻠﻰ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ﺍﳌﻘﺎﺑﻞ ﺇﻥ ﲢﻘﻖ ﺍﻟﺸﺮﻁ‪ ،‬ﺃﻭ ﻋﻠﻰ ﺍﳌﺆﺷﺮ ﺍﳌﻌﺪﻝ ﰲ ﺍﳊﺎﻟﺔ‬
‫ﺍﻷﺧﺮﻯ‪.‬‬
‫‪y ij‬‬
‫= ‪y aij‬‬
‫‪sj‬‬
‫‪ -2‬ﺍﳊﺎﻟﺔ ﺍﻟﺘﺠﻤﻴﻌﻴﺔ‪:‬‬
‫ﻟﺘﻔﺎﺩﻱ ﺍﻟﺘﻜﺮﺍﺭ ﻧﻘﺘﺼﺮ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻋﻠﻰ ﺫﻛﺮ ﺍﳌﺮﺍﺣﻞ ﻓﻘﻂ ﻭ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y ij = L ij + C ij + S j + I ij‬‬ ‫)‪(4.20‬‬

‫~‬
‫‪a) y ij = L ij + C ij‬‬
‫~‬ ‫~‬
‫‪b) z ij = y ij - y ij‬‬
‫~‬
‫‪c) z ij = S j + I ij‬‬
‫‪1 m‬‬
‫= ‪d) S j‬‬ ‫‪å z ij‬‬
‫‪m - 1 i =1‬‬
‫‪p‬‬
‫‪e) å S j = 0‬‬
‫‪j=1‬‬
‫_‬
‫‪sj = S j - S‬‬ ‫ﺇﺫﺍ ﱂ ﻳﺘﺤﻘﻖ ﺍﻟﺸﺮﻁ‬
‫‪yija = yij - s j‬‬
‫ﻣﺜﺎﻝ ‪ :22‬ﺑﻌﺪ ﻛﺸﻒ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻄﺮﻳﻘﺔ )‪ ،(KW‬ﻭﺑﺎﺳﺘﻌﻤﺎﻝ ﻧﻔﺲ‬
‫ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﻮﺍﺭﺩﺓ ﰲ ﺍﳉﺪﻭﻝ)‪ (10‬ﺻﻔﺤﺔ )‪ ،(36‬ﺍﳌﻄﻠﻮﺏ ﺇﺯﺍﻟﺘﻬﺎ ‪‬ﺬﻩ‬
‫ﺍﻟﻄﺮﻳﻘﺔ‪.‬‬

‫‪114‬‬
 ‫ﺍﳉﺪﻭﻝ )‪ :(24‬ﲢﺪﻳﺪ ﺍﻟﻌﻼﻗﺔ ﺑﲔ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫‪-‬‬ ‫‪-‬‬ ‫‪-‬‬
‫‪yi‬‬
‫‪2‬‬
‫‪s i yi‬‬ ‫‪si‬‬ ‫‪yi‬‬ ‫‪.4‬‬ ‫‪.3‬‬ ‫‪.2‬‬ ‫‪.1‬‬ ‫ﺍﻟﺴﻨﺔ‪/‬ﺍﻟﻔﺼﻞ‬
‫‪612.56‬‬ ‫‪282.89‬‬ ‫‪11.43‬‬ ‫‪24.75‬‬ ‫‪21‬‬ ‫‪44‬‬ ‫‪20‬‬ ‫‪14‬‬ ‫‪1988‬‬
‫‪976.56‬‬ ‫‪639.37‬‬ ‫‪20.46‬‬ ‫‪31.25‬‬ ‫‪32‬‬ ‫‪64‬‬ ‫‪19‬‬ ‫‪10‬‬ ‫‪1989‬‬
‫‪915.06‬‬ ‫‪691.82‬‬ ‫‪22.87‬‬ ‫‪30.25‬‬ ‫‪29‬‬ ‫‪68‬‬ ‫‪12‬‬ ‫‪12‬‬ ‫‪1990‬‬
‫‪915.06‬‬ ‫‪606.51‬‬ ‫‪20.05‬‬ ‫‪30.25‬‬ ‫‪36‬‬ ‫‪60‬‬ ‫‪18‬‬ ‫‪7‬‬ ‫‪1991‬‬
‫‪107.56‬‬ ‫‪812.53‬‬ ‫‪24.81‬‬ ‫‪32.75‬‬ ‫‪50‬‬ ‫‪64‬‬ ‫‪11‬‬ ‫‪6‬‬ ‫‪1992‬‬
‫‪4491.8‬‬ ‫‪3033.1‬‬ ‫‪99.62‬‬ ‫‪149.25‬‬ ‫اﻟﻤﺠﺎﻣﯿﻊ‬

‫ﺇﺿﺎﻓﺔ ﺍﻷﺭﺑﻌﺔ ﺃﻋﻤﺪﺓ ﺍﻷﺧﲑﺓ ﰲ ﺍﳉﺪﻭﻝ )‪ ،(24‬ﺟﺎﺀ ﻟﻐﺮﺽ ﺍﻟﺒﺤﺚ ﻋﻦ‬

‫ﺍﻟﻌﻼﻗﺔ ﺍﳉﺎﻣﻌﺔ ﺑﲔ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ‪ ،‬ﻭﻳﺘﻢ ﺫﻟﻚ ﺑﺘﻘﺪﻳﺮ ﺍﳌﻌﺎﺩﻟﺔ ﺃﺩﻧﺎﻩ‪:‬‬
‫_‬
‫‪si = a+ by‬‬ ‫‪i‬‬
‫‪m‬‬
‫_‬
‫‪1‬‬ ‫‪99. 62‬‬
‫=‪s‬‬ ‫‪ås‬‬ ‫‪i‬‬ ‫=‬ ‫‪= 19. 92‬‬ ‫ﺃﻳﻦ‬
‫‪m‬‬ ‫‪i=1‬‬ ‫‪5‬‬
‫‪m‬‬
‫=‬ ‫‪1‬‬ ‫_‬
‫‪149. 25‬‬
‫=‪y‬‬
‫‪m‬‬
‫= ‪åyi‬‬
‫‪i =1‬‬ ‫‪5‬‬
‫‪= 28. 85‬‬
‫‪m‬‬ ‫_‬ ‫_‬ ‫=‬ ‫‪m‬‬ ‫_‬ ‫= _‬

‫)‪å (s i - s )( yi - y‬‬ ‫‪ås i yi - ms y‬‬

‫= ˆ‪b‬‬ ‫‪i =1‬‬
‫‪m‬‬ ‫_‬ ‫=‬
‫=‬ ‫‪i =1‬‬
‫‪m‬‬ ‫‪-‬‬ ‫=‬
‫‪å ( yi - y) 2‬‬
‫‪i =1‬‬
‫‪åy‬‬ ‫‪i‬‬
‫‪2‬‬
‫‪-my‬‬ ‫‪2‬‬

‫‪i =1‬‬
‫)‬
‫ﺏ‪b = 1.62‬‬ ‫ﺑﺎﻟﺘﻌﻮﻳﺾ ﰲ ﻋﻼﻗﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﺃﻋﻼﻩ‪ ،‬ﻗﺪﺭﺕ ﺍﳌﻌﻠﻤﺔ‬
‫)‬
‫ﺣﻴﺚ ‪ b > 0.10‬ﻭﺍﻟﱵ ﺗﻌﲏ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺫﺍﺕ ﻋﻼﻗﺔ ﺟﺪﺍﺋﻴﺔ ﺑﲔ ﺍﳌﺮﻛﺒﺎﺕ‪.‬‬
‫ﺗﺘﻢ ﺍﻵﻥ ﻋﻤﻠﻴﺔ ﺇﺯﺍﻟﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻭﻓﻖ ﺍﻟﺸﻜﻞ ﺍﳉﺪﺍﺋﻲ ﻭﺣﺴﺐ ﺍﳌﺮﺍﺣﻞ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫ﺍ ‪ -‬ﺗﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﳌﻤﺮﻛﺰﺓ ﻣﻊ ﺍﻓﺘﺮﺍﺽ ﺃﻥ )‪:(n=p‬‬
‫~‬ ‫‪1‬‬ ‫‪1‬‬ ‫‪1‬‬
‫= ‪yt‬‬ ‫(‬ ‫‪2‬‬
‫‪y t + 2 + y t +1 + y t + y t -1 +‬‬ ‫‪2‬‬
‫) ‪y t-2‬‬
‫‪4‬‬
‫ﻳﻔﻘﺪ ﺗﻄﺒﻴﻖ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻣﺸﺎﻫﺪﺗﲔ ﰲ ﺑﺪﺍﻳﺔ ﻭ‪‬ﺎﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﻛﻤﺎ ﻳﺘﺒﲔ ﰲ‬
‫ﺍﳉﺪﻭﻝ ﺍﻟﺘﺎﱄ )ﺑﻌﺪ ﺍﻻﻧﺘﻘﺎﻝ ﻣﻦ ﺍﻟﺘﺮﺗﻴﺐ ﺍﻟﺰﻣﲏ ﺇﱃ ﺍﳉﺪﻭﱄ(‪:‬‬

‫‪115‬‬
 ‫ﺍﳉﺪﻭﻝ )‪ :(25‬ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﻤﻬﺪﺓ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﳌﻤﺮﻛﺰﺓ‬
‫‪.4‬‬ ‫‪.3‬‬ ‫‪.2‬‬ ‫‪.1‬‬ ‫ﺍﻟﺴﻨﺔ‪/‬ﺍﻟﻔﺼﻞ‬
‫‪23.63‬‬ ‫‪24.25‬‬ ‫‪-‬‬ ‫‪-‬‬ ‫‪1988‬‬
‫‪30.88‬‬ ‫‪31.50‬‬ ‫‪29.88‬‬ ‫‪26‬‬ ‫‪1989‬‬
‫‪29.75‬‬ ‫‪29.63‬‬ ‫‪30.63‬‬ ‫‪30.50‬‬ ‫‪1990‬‬
‫‪29.13‬‬ ‫‪13.30‬‬ ‫‪29.38‬‬ ‫‪29.50‬‬ ‫‪1991‬‬
‫‪-‬‬ ‫‪-‬‬ ‫‪31‬‬ ‫‪28.75‬‬ ‫‪1992‬‬

‫~‬
‫‪y ij = L ij . C ij‬‬ ‫ﺃﻱ‬
‫ﺏ‪ -‬ﺗﻘﺴﻴﻢ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ ﻋﻠﻰ ﺍﻟﻨﺎﲡﺔ ﺃﻋﻼﻩ‪ ،‬ﺑﻐﻴﺔ ﺍﻟﺘﺨﻠﺺ ﻣﻦ‬
‫‪yij‬‬
‫~ = ‪zij‬‬ ‫ﺍﳌﺮﻛﺒﺘﲔ ﺍﻟﻨﻈﺎﻣﻴﺘﲔ‬
‫‪yij‬‬

‫ﺍﳉﺪﻭﻝ )‪ :(26‬ﺍﻟﺘﺨﻠﺺ ﻣﻦ ﺍﳌﺮﻛﺒﺎﺕ ﺍﻟﻨﻈﺎﻣﻴﺔ‬

‫ﺍﻟﻔﺼل‬
‫‪.4‬‬ ‫‪.3‬‬ ‫‪.2‬‬ ‫‪.1‬‬
‫ﺍﻟﺴﻨﺔ‬
‫‪0.89‬‬ ‫‪1.81‬‬ ‫‪-‬‬ ‫‪-‬‬ ‫‪1988‬‬
‫‪1.04‬‬ ‫‪2.03‬‬ ‫‪0.64‬‬ ‫‪0.38‬‬ ‫‪1989‬‬
‫‪0.97‬‬ ‫‪2.29‬‬ ‫‪0.39‬‬ ‫‪0.39‬‬ ‫‪1990‬‬
‫‪1.24‬‬ ‫‪1.99‬‬ ‫‪0.61‬‬ ‫‪0.24‬‬ ‫‪1991‬‬
‫‪-‬‬ ‫‪-‬‬ ‫‪0.35‬‬ ‫‪0.21‬‬ ‫‪1992‬‬

‫ﺟـ ‪ -‬ﺣﺴﺎﺏ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﻭﻫﻲ ﻋﺒﺎﺭﺓ ﻋﻦ ﻣﺘﻮﺳﻂ ﳎﻤﻮﻉ‬

‫ﻋﻨﺎﺻﺮ ﻛﻞ ﻋﻤﻮﺩ ﻣﻘﺎﺑﻞ ﻟﻔﺼﻞ ﻣﻌﲔ‪ ،‬ﺃﻱ‪:‬‬
‫‪1 5‬‬
‫= ‪Sj‬‬ ‫‪å z ij‬‬
‫‪4 i =1‬‬

‫) ‪S 1 = ( 0. 38 + 0. 39 + 0. 24 + 0. 21‬‬ ‫ﻣﻨﻪ‬
‫ﺑﺸﻜﻞ ﻋﺎﻡ ﺗﻜﻮﻥ ﺍﻟﻨﺘﺎﺋﺞ ﻛﻤﺎ ﻫﻲ ﻣﻮﺿﺤﺔ ﰲ ﺍﻟﺸﻜﻞ ﺍﳉﺪﻭﻝ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪116‬‬
 ‫ﺍﳉﺪﻭﻝ )‪ :(27‬ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ‬
‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫ﺍﳌﺆﺷﺮ‪/‬ﺍﻟﻔﺼﻞ‬
‫‪1.04‬‬ ‫‪2.03‬‬ ‫‪0.5‬‬ ‫‪0.31‬‬ ‫‪S j‬‬
‫‪4‬‬

‫‪åS‬‬ ‫‪j‬‬ ‫‪= 3. 88‬‬ ‫‪:4‬‬

‫ﺩ ‪ -‬ﻛﻮﻥ ﺍﻟﺸﺮﻁ ﱂ ﻳﺘﺤﻘﻖ ﺃﻱ‬
‫‪j= 1‬‬
‫ﻧﻌﻤﻞ ﻋﻠﻰ ﺣﺴﺎﺏ ﺍﳌﺆﺷﺮﺍﺕ ﺍﳌﻌﺪﻟﺔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪Sj‬‬
‫= ‪sj‬‬ ‫_‬
‫‪S‬‬
‫‪3. 88‬‬
‫= ‪S‬‬ ‫‪= 0. 97‬‬ ‫ﺣﻴﺚ‬
‫‪4‬‬
‫‪0. 31‬‬
‫= ‪s1‬‬ ‫‪= 0. 32‬‬ ‫ﻣﻨﻪ‬
‫‪0. 97‬‬
‫ﻭﻫﻜﺬﺍ‪.‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(28‬ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ ﺍﳌﻌﺪﻟﺔ‬
‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫ﺍﳌﺆﺷﺮ‪/‬ﺍﻟﻔﺼﻞ‬
‫‪1.07‬‬ ‫‪2.09‬‬ ‫‪0.525‬‬ ‫‪sj‬‬ ‫‪0.32‬‬

‫ﻫـ‪ -‬ﺣﺴﺎﺏ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﻭﺫﻟﻚ ﺑﻘﺴﻤﺔ ﺍﻟﺴﻠﺴﻠﺔ‬

‫ﺍﻷﺻﻠﻴﺔ ﻋﻠﻰ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ﺍﳌﻘﺎﺑﻞ‬
‫‪yij‬‬
‫= ‪y a ij‬‬
‫‪sj‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(29‬ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ‬
‫‪.4‬‬ ‫‪.3‬‬ ‫‪.2‬‬ ‫‪.1‬‬ ‫ﺍﻟﺴﻨﺔ‪/‬ﺍﻟﻔﺼﻞ‬
‫‪19.63‬‬ ‫‪21.05‬‬ ‫‪38.46‬‬ ‫‪43.75‬‬ ‫‪1988‬‬
‫‪29.91‬‬ ‫‪30.62‬‬ ‫‪36.54‬‬ ‫‪25.31‬‬ ‫‪1989‬‬
‫‪27.10‬‬ ‫‪32.54‬‬ ‫‪23.08‬‬ ‫‪37.5‬‬ ‫‪1990‬‬
‫‪33.64‬‬ ‫‪28.71‬‬ ‫‪34.62‬‬ ‫‪21.88‬‬ ‫‪1991‬‬
‫‪46.73‬‬ ‫‪30.62‬‬ ‫‪21.15‬‬ ‫‪18.75‬‬ ‫‪1992‬‬

‫‪ -4‬ﺣﱴ ﺗﻜﻮﻥ ﺍﳊﺴﺎﺑﺎﺕ ﺩﻗﻴﻘﺔ ﳚﺐ ﺃﺧﺬ ﺃﻛﺜﺮ ﻣﻦ ﺭﻗﻤﲔ ﺑﻌﺪ ﺍﻟﻔﺎﺻﻠﺔ‪.‬‬

‫‪117‬‬
 ‫ﺍﻟﻄﺮﻳﻘﺔ ﺍﻻﳓﺪﺍﺭﻳﺔ‪:‬‬
‫ﻟﺘﺒﺴﻴﻂ ﺍﻟﻔﻜﺮﺓ‪ ،‬ﻧﻔﺘﺮﺽ ﺗﻮﻓﺮ ﺳﻠﺴﻠﺔ ﺯﻣﻨﻴﺔ ﺫﺍﺕ ﻣﺮﻛﺒﺔ ﻓﺼﻠﻴﺔ ﺑﺪﻭﺭﻳﺔ‬
‫ﻣﻘﺪﺍﺭﻫﺎ ‪ p = 4‬ﻭﻋﺸﻮﺍﺋﻴﺔ ﻓﻘﻂ ﻭﰲ ﺷﻜﻞ ﲡﻤﻴﻌﻲ‪ .‬ﻓﻔﻲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻧﻌﱪ ﻋﻦ‬
‫ﺍﻟﻔﺼﻠﻴﺔ ﺑﺎﳌﺘﻐﲑﺍﺕ ﺍﻟﺘﻤﺜﻴﻠﻴﺔ )‪ (Dummy variables‬ﻭﻛﺎﻵﰐ‪:‬‬

‫‪D1t = 1‬‬‫ﺇﺫﺍ ﻛﺎﻧﺖ ﺍﳌﺸﺎﻫﺪﺓ ﺧﺎﺻﺔ ﺑﺎﻟﻔﺼﻞ ﺍﻷﻭﻝ‬

‫‪=0‬‬ ‫ﺑﺎﻟﻔﺼﻮﻝ ﺍﻷﺧﺮﻯ‬
‫ﺇﺫﺍ ﻛﺎﻧﺖ ﺍﳌﺸﺎﻫﺪﺓ ﺧﺎﺻﺔ ﺑﺎﻟﻔﺼﻞ ﺍﻟﺜﺎﱐ ‪D2t = 1‬‬
‫‪=0‬‬ ‫ﺑﺎﻟﻔﺼﻮﻝ ﺍﻷﺧﺮﻯ‬
‫ﺇﺫﺍ ﻛﺎﻧﺖ ﺍﳌﺸﺎﻫﺪﺓ ﺧﺎﺻﺔ ﺑﺎﻟﻔﺼﻞ ﺍﻟﺜﺎﻟﺚ ‪D3t = 1‬‬
‫‪=0‬‬ ‫ﺑﺎﻟﻔﺼﻮﻝ ﺍﻷﺧﺮﻯ‬
‫ﺇﺫﺍ ﻛﺎﻧﺖ ﺍﳌﺸﺎﻫﺪﺓ ﺧﺎﺻﺔ ﺑﺎﻟﻔﺼﻞ ﺍﻟﺮﺍﺑﻊ ‪D4t = 1‬‬
‫‪=0‬‬ ‫ﺑﺎﻟﻔﺼﻮﻝ ﺍﻷﺧﺮﻯ‬
‫ﳝﻜﻦ ﳕﺬﺟﺔ ﺍﻟﻔﺼﻠﻴﺔ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﻌﺎﻡ ﺍﻟﺘﺎﱄ ‪:‬‬
‫‪5‬‬

‫‪p- 1‬‬
‫‪y t = a + å g j D jt + u t‬‬ ‫)‪(4.21‬‬
‫‪j=1‬‬
‫ﰲ ﺍﳌﺜﺎﻝ ﺃﻋﻼﻩ ﻳﺼﺒﺢ ﺍﻟﻨﻤﻮﺫﺝ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪3‬‬
‫‪yt = a + åg j D jt + ut‬‬ ‫)‪(4.22‬‬
‫‪j =1‬‬

‫‪y t = a + g 1 D1t + g 2 D2 t + g 3 D3t + u t‬‬ ‫ﺃﻱ‬

‫ﻓﻨﻼﺣﻆ ﺟﻴﺪﺍﹰ ﻏﻴﺎﺏ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ﺍﻟﺮﺍﺑﻊ ﺑﺴﺒﺐ ﺩﺧﻮﻝ ﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ ‪a‬‬
‫ﰲ ﺍﳌﻌﺎﺩﻟﺔ ﻭﺍﻟﺬﻱ ﻻ ﻳﺴﻤﺢ ﺑﺈﺩﺭﺍﺝ ﻣﺘﻐﲑ ﲤﺜﻴﻠﻲ ﺇﺿﺎﰲ ﳌﺼﻔﻮﻓﺔ ﺍﳌﺘﻐﲑﺍﺕ‬

‫‪ -5‬ﺑﻴﻨﻤﺎ ﰲ ﺍﳊﺎﻟﺔ ﺍﳉﺪﺍﺋﻴﺔ ﳝﻜﻦ ﻛﺘﺎﺑﺘﻬﺎ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‬

‫‪y t = a . Õ pj =-11 g j D jt :‬‬

‫‪118‬‬
‫ﺍﻟﺸﺎﺭﺣﺔ ﻭﺍﻟﱵ ﻧﺴﻤﻴﻬﺎ ﰲ ﺍﻟﻐﺎﻟﺐ ﺏ ‪ ، X‬ﻛﻮﻧﻪ ﳚﻌﻞ ﺍﻟﻌﻤﻮﺩ ﺍﻷﻭﻝ )ﺍﻷﺣﺎﺩﻱ‬
‫‪6‬‬

‫ﺍﻟﻌﻨﺎﺻﺮ( ﻳﺴﺎﻭﻱ ﺷﻌﺎﻉ ‪‬ﻤﻮﻉ ﻋﻨﺎﺻﺮ ﺍﻷﻋﻤﺪﺓ ﺍﻷﺭﺑﻌﺔ ﺍﻷﺧﺮﻯ )ﺍﻷﺣﺎﺩﻳﺔ‬

‫ﺍﻟﻌﻨﺎﺻﺮ(‪ ،‬ﻭﻫﺬﺍ ﻳﺘﺮﻙ ﺍﳌﺼﻔﻮﻓﺔ ﻫﺬﻩ ﻧﺎﻗﺼﺔ ﺍﻟﺮﺗﺒﺔ ﻭﺑﺎﻟﺘﺎﱄ ﻓﺈﻥ ﺍﳌﺼﻔﻮﻓﺔ‬
‫) ‪ ( X X‬ﺗﻜﻮﻥ ﻛﺬﻟﻚ ﺃﻳﻀﺎﹰ‪ ،‬ﺫﺍﺕ ﳏﺪﺩ ﻣﻌﺪﻭﻡ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﻻ ﳝﻜﻦ ﻗﻠﺒﻬﺎ ﻣﻦ‬
‫\‬

‫ﺃﺟﻞ ﺍﳊﻞ ﺍﻟﻮﺣﻴﺪ )‪ .(Unique solution‬ﻟﻠﺤﺼﻮﻝ ﻋﻠﻰ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ) ‪( p‬‬

‫ﺍﻟﺮﺍﺑﻊ ﻫﻨﺎ‪ ،‬ﻧﺴﺘﻌﲔ ﺑﺸﺮﻁ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻟﺴﺎﺑﻘﺔ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪ (1‬ﺍﳊﺎﻟﺔ ﺍﻟﺘﺠﻤﻴﻌﻴﺔ‬
‫‪p-1‬‬

‫‪åS‬‬
‫‪j= 1‬‬
‫‪j‬‬ ‫‪+S‬‬ ‫‪p‬‬ ‫‪=0‬‬
‫‪p- 1‬‬
‫‪S p = -å S j‬‬ ‫ﻣﻨﻪ‬
‫‪j=1‬‬
‫ﺣﻴﺚ ‪ S p‬ﳝﺜﻞ ‪. g p‬‬
‫‪p- 1‬‬
‫‪S‬‬ ‫‪p‬‬ ‫‪= p-åS‬‬ ‫‪j‬‬ ‫‪ (2‬ﺍﳊﺎﻟﺔ ﺍﳉﺪﺍﺋﻴﺔ‬
‫‪j= 1‬‬

‫ﻣﺜﺎﻝ ‪ :23‬ﻟﺘﻮﺿﻴﺢ ﺫﻟﻚ ﻧﺘﺬﻛﺮ ﺍﳌﺜﺎﻝ )‪ (10‬ﻣﻊ ﻧﻔﺲ ﺍﳌﻌﻄﻴﺎﺕ‪ ،‬ﺃﻳﻦ ﰎﹶ ﺍﻟﺘﻮﺻﻞ‬
‫ﺃﺛﻨﺎﺀ ﺍﳊﻞ ﺇﱃ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﺪﺭﻭﺳﺔ ﺫﺍﺕ ﻣﺮﻛﺒﺔ ﻓﺼﻠﻴﺔ ﺑﻌﻼﻗﺔ ﺟﺪﺍﺋﻴﺔ ﻣﻊ ﻏﻴﺎﺏ‬
‫ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﺇﺫﹰﺍ ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻨﻬﺎ ﲟﺎ ﻳﻠﻲ‪:‬‬
‫‪w t = ag 1D1 t .g 2D 2 t .g 3D3 t .e u t‬‬ ‫)‪(4.23‬‬

‫ﺃﻳﻦ ﺃﺩﺭﺟﻨﺎ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﰲ ﺷﻜﻞ ﺃﺳﻲ ﻟﺘﺴﻬﻴﻞ ﺍﻟﺘﺤﻮﻳﻞ ﺍﻟﻠﻮﻏﺎﺭﲤﻲ‬

‫) ‪u ® N( 0, s‬‬
‫‪t‬‬
‫ﻭﺿﻤﺎﻥ ﺍﺳﺘﻤﺮﺍﺭ ﻓﺮﺿﻴﺔ ﺍﻟﺘﻮﺯﻳﻊ ﺍﻟﻄﺒﻴﻌﻲ ﻟﻸﺧﻄﺎﺀ ﺃﻱ‪:‬‬
‫‪2‬‬

‫ﺇﺫﹰﺍ ﺑﻌﺪ ﺍﻟﺘﺤﻮﻳﻞ‬

‫‪ln(wt ) = lna + ln g 1 D1t + ln g 2 D2t + ln g 3 D3t + ut‬‬ ‫)‪(4.24‬‬

‫)‪.(52‬‬ ‫‪ -6‬ﻟﺸﻜﻞ ﻣﺸﺎﺑﻪ‪ ،‬ﺃﻧﻈﺮ ﺍﻟﻌﻼﻗﺔ )‪ (3.5‬ﺻﻔﺤﺔ‬

‫‪119‬‬
 ‫ﺇﻥ ﺍﻟﺘﻘﺪﻳﺮ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﺍﻟﻌﺎﺩﻳﺔ ﻳﻌﻄﻲ ﺍﻟﻨﺘﺎﺋﺞ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫‪7‬‬

‫)‬
‫‪ln(wt ) = 3.29 - 1.098D1t - 0.55D2t + 0.67D3t‬‬
‫)‪(0.08) (0.118‬‬ ‫)‪(0.12‬‬ ‫)‪(0.12‬‬

‫ﺑﺎﺳﺘﺨﺪﺍﻡ ﺇﺣﺼﺎﺀﺓ ﺳﺘﻴﻮﺩﻧﺖ ﳒﺪ ﺃﻥ ﻛﻞ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ ﺫﺍﺕ ﺩﻻﻟﺔ‬

‫ﺇﺣﺼﺎﺋﻴﺔ‪ ،‬ﻭﻛﻮﻥ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻌﺪ ﺍﺳﺘﺨﺪﺍﻡ ﺍﻟﻌﻼﻗﺔ ﺍﻷﺳﻴﺔ ﻫﻲ‪:‬‬
‫‪S 1 = exp( -1. 098 ) = 0. 33‬‬

‫‪S 2 = 0. 57‬‬
‫‪. S 3 = 1. 95‬‬
‫ﻧﺴﺘﺨﺮﺝ ﺍﳌﺆﺷﺮ ﺍﻷﺧﲑ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺸﺮﻃﻴﺔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪p- 1‬‬
‫‪S 4 = 4 - å S j = 4 - 2. 85 = 1. 15‬‬
‫‪j= 1‬‬

‫ﻫﺬﺍ ﰲ ﺣﺎﻟﺔ ﻛﻮﻥ ﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ ﻣﻌﻨﻮﻱ ﻛﻤﺎ ﻫﻮ ﺍﳊﺎﻝ ﰲ ﻫﺬﺍ ﺍﳌﺜﺎﻝ‪ ،‬ﺑﻴﻨﻤﺎ‬
‫ﰲ ﺣﺎﻟﺔ ﺍﻟﻌﻜﺲ‪ ،‬ﻳﺰﺍﻝ ﻫﺬﺍ ﺍﻷﺧﲑ ﻣﻊ ﺇﺿﺎﻓﺔ ﺍﳌﺘﻐﲑ ﺍﻟﺘﻤﺜﻴﻠﻲ ) ‪ ( p‬ﺍﻟﺮﺍﺑﻊ ﻫﻨﺎ‪ ،‬ﻣﻊ‬
‫ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ ﺍﳌﻌﲏ ﺃﻱ‪:‬‬
‫‪4‬‬
‫‪y t = å g j D jt + u t‬‬ ‫)‪(4.25‬‬
‫‪j= 1‬‬

‫ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻘﺪﻳﺮ ﺑﺎﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﻟﻠﺤﺼﻮﻝ ﻋﻠﻰ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ‬

‫‪ g‬ﺍﳌﻤﺜﻠﺔ ﻟـ ‪ . S‬ﺗﻜﻮﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﻟﻔﺼﻠﻴﺔ ﻫﻲ ﻋﺒﺎﺭﺓ ﻋﻦ ﺍﻟﺒﻮﺍﻗﻲ‬
‫ﺍﻟﻨﺎﲡﺔ ﻣﻦ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻘﺪﻳﺮ ﺃﻱ‪:‬‬
‫‪et = y ta = y t - yˆ t‬‬ ‫)‪(4.26‬‬

‫‪ -7‬ﻣﺎ ﺑﲔ ﺍﻟﻘﻮﺳﲔ ﳝﺜﻞ ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ ﻟﻠﻤﻌﻠﻤﺔ ﺃﻋﻼﻩ ‪.‬‬

‫‪120‬‬
 ‫ﻛﻤﺎ ﳝﻜﻦ ﺣﺴﺎ‪‬ﺎ ﺑﺎﻟﻄﺮﻕ ﺍﻟﺴﺎﻟﻔﺔ ﺍﻟﺬﻛﺮ‪.‬‬
‫ﺗﻜﻮﻥ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﺑﻌﺪ ﺇﺯﺍﻟﺔ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﻣﺘﻤﺜﻠﺔ ﰲ ﺇﺿﺎﻓﺔ ﺍﳌﺆﺷﺮ ﺍﻟﻔﺼﻠﻲ‬
‫ﺍﳌﻘﺎﺑﻞ )ﺍﳊﺎﻟﺔ ﺍﻟﺘﺠﻤﻴﻌﻴﺔ( ﺃﻭ ﺑﻀﺮﺑﻪ ﰲ ﺍﳌﺸﺎﻫﺪﺓ ﺍﳌﻌﺪ‪‬ﻟﺔ )ﺍﳉﺪﺍﺋﻴﺔ( ﰲ ﺍﻟﻔﺘﺮﺓ‬
‫)‪ (T+L‬ﺍﻟﱵ ﺗﻜﻮﻥ ﻗﺪ ﰎﹶ ﺍﳊﺼﻮﻝ ﻋﻠﻴﻬﺎ ﺑﻄﺮﻳﻘﺔ ﺗﻨﺒﺆﻳﺔ ﻣﻼﺋﻤﺔ ﺣﺴﺐ ﺍﻟﻌﻼﻗﺔ‬
‫ﺍﳌﻮﺟﻮﺩﺓ‪ ،‬ﻓﻔﻲ ﺍﳊﺎﻟﺔ ﺍﻟﺘﺠﻤﻴﻌﻴﺔ ﻣﺜﻼﹰّ‪ ،‬ﺗﻜﺘﺐ‪:‬‬
‫‪yˆ T + L, j = yˆ Ta + L, j + S j‬‬ ‫)‪(4.27‬‬

‫ﻫﺬﺍ ﺣﺎﻝ ﺍﻟﺘﻌﺎﻣﻞ ﻣﻊ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻄﺮﻳﻘﺔ ﺍﻟﻌﺰﻝ ﺃﻱ ﺑﻄﺮﻳﻘﺔ ﻏﲑ ﻣﺒﺎﺷﺮﺓ ﺣﻴﺚ‬

‫ﺃ‪‬ﺎ ﻻ ﺗﺘﻌﺎﻣﻞ ﻣﻊ ﻛﻞ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺁﻧﻴﺎﹰ‪.‬‬

‫‪ -2‬ﺍﻟﻨﻤﺬﺟﺔ‪:‬‬
‫ﺿﻤﻦ ﳕﺎﺫﺝ ﺍﻻﺳﺘﻘﻄﺎﺏ‪ ،‬ﳝﻜﻦ ﳕﺬﺟﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺑﺸﻜﻞ ﻣﺒﺎﺷﺮ‪ ،‬ﻫﺬﺍ ﻳﺴﻤﺢ‬
‫ﻃﺒﻌ‪‬ﺎ ﺑﻜﺴﺐ ﻣﻌﻠﻮﻣﺎﺕ ﺇﺿﺎﻓﻴﺔ ﻣﻘﺎﺑﻞ ﺍﻟﺼﻌﻮﺑﺔ ﺍﻟﱵ ﺗﻮﺍﺟﻬﻨﺎ ﻭﺍﻟﱵ ﺗﻼﺷﺖ‬
‫ﺑﺘﻮﺍﺟﺪ ﺃﺟﻬﺰﺓ ﺍﻟﻜﻤﺒﻴﻮﺗﺮ ﺍﻟﺴﺮﻳﻌﺔ ﺟﺪﺍﹰ‪ .‬ﻧﻘﺘﺼﺮ ﰲ ﻫﺬﺍ ﺍ‪‬ﺎﻝ ﻋﻠﻰ ﺗﻨﺎﻭﻝ ﺍﻟﻄﺮﻳﻘﺔ‬
‫ﺍﳌﻮﺍﻟﻴﺔ ﻝﻫﻮﻟﺖ ﻭ ﻭﻧﺘﺮﺯ )‪ ،(Holt-winters‬ﻭﻻ ﻧﺘﻄﺮﻕ ﺇﱃ ﻃﺮﻳﻘﺔ ‪(Buys-‬‬
‫)‪ Ballot‬ﻛﻮ‪‬ﺎ ﺗﺴﺘﻌﻤﻞ ﻓﻘﻂ ﰲ ﺣﺎﻝ ﺍﺣﺘﻮﺍﺀ ﺍﻟﺴﻠﺴﻠﺔ ﻋﻠﻰ ﺍﳌﺮﻛﺒﺎﺕ ﺍﻟﺜﻼﺙ‬
‫ﺑﻌﻼﻗﺔ ﲡﻤﻴﻌﻴﺔ ﻓﻘﻂ‪ .‬ﻣﻦ ﺍﻟﻄﺮﻕ ﺍﻟﺸﻬﲑﺓ ﰲ ﻫﺬﺍ ﺍ‪‬ﺎﻝ‪ ،‬ﻧﺬﻛﺮ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫)‪:(Holt-Winters‬‬ ‫ﻃﺮﻳﻘﺔ ﻫﻮﻟﺖ‪-‬ﻭﻧﺘﺮﺯ‬

‫ﺗﻌﻜﺲ ﺍﻟﻄﺮﻳﻘﺔ ﻣﺴﺎﳘﺔ )‪ (Winters‬ﻣﻦ ﺧﻼﻝ ﺇﺿﺎﻓﺔ ﺇﱃ ﻣﻌﺎﺩﻟﱵ ﻫﻮﻟﺖ‬
‫)‪ (Holt‬ﺗﻠﻚ ﺍﳋﺎﺻﺔ ﺑﺎﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﻭﳝﻜﻦ ﻛﺘﺎﺑﺔ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳉﺪﻳﺪ‬
‫ﻭﺑﺸﻜﻞ ﻣﻦ ﺍﻟﺘﺤﻮﻳﺮ ﺣﱴ ﻳﺘﺠﺎﻭﺏ ﻣﻊ ﺍﳌﺮﻛﺒﺎﺕ ﺍﻟﺜﻼﺛﺔ ﺁﻧﻴﺎﹰ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪121‬‬
 ‫~‬ ‫‪a‬‬
‫) ‪y t = a ( yt ) + (1 - a )( ~yt -1 + rt -1‬‬

‫)‪(4.28‬‬
‫~ ( = ‪rt‬‬
‫~ ‪yt -‬‬
‫‪y t -1 ) + (1 - g ) rt -1‬‬

‫‪S t = b ( z t ) + (1 - b ) S t - p‬‬

‫ﺣﻴﺚ ﰲ ﺍﳊﺎﻟﺔ ﺍﻟﺘﺠﻤﻴﻌﻴﺔ‬

‫‪y at = y t - S t - p‬‬
‫~‬
‫‪z t = yt - yt‬‬
‫ﺑﻴﻨﻤﺎ ﰲ ﺍﳊﺎﻟﺔ ﺍﳉﺪﺍﺋﻴﺔ‬
‫‪yt‬‬
‫= ‪y at‬‬
‫‪S t-p‬‬
‫‪yt‬‬
‫= ‪zt‬‬ ‫~‬
‫‪yt‬‬
‫ﺗﺘﻤﻴﺰ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﰲ ﻛﻮ‪‬ﺎ ﺫﺍﺗﻴﺔ ﺍﻟﺘﻌﺪﻳﻞ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﻻ ﺗﻜﻮﻥ ﻣﺆﺷﺮﺍ‪‬ﺎ‬
‫ﺍﻟﻔﺼﻠﻴﺔ ﺑﺎﻟﻀﺮﻭﺭﺓ ﻣﺴﺎﻭﻳﺔ ﳌﺆﺷﺮﺍﺕ ﺍﻟﺴﻨﺔ ﺍﻟﱵ ﻗﺒﻠﻬﺎ ﺃﻭ ﺑﻌﺪﻫﺎ‪.‬‬
‫ﻋﻠﻰ ﺃﺳﺎﺱ ﺗﺼﻐﲑ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ‪å e‬‬
‫‪2‬‬
‫‪t‬‬ ‫ﻳﺘﻢ ﺣﺴﺎﺏ‬ ‫‪g , b, a‬‬
‫ﺃﻳﻦ ˆ‪ e = y - y‬ﻛﻤﺎ ﲢﺴﺐ ‪ ŷt‬ﰲ ﺍﳊﺎﻟﺘﲔ ﺍﳉﺪﺍﺋﻴﺔ ﻭﺍﻟﺘﺠﻤﻴﻌﻴﺔ ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪:‬‬ ‫‪t‬‬ ‫‪t‬‬ ‫‪t‬‬

‫~‬
‫‪$y t = ( y t + rt ) S t - p‬‬ ‫)‪(4.29‬‬
‫~‬
‫‪$y t = ( y t + rt ) + S t - p‬‬ ‫)‪(4.30‬‬

‫‪t = ( p + 1 ), T‬‬
‫ﻷﳘﻴﺘﻬﺎ‪ ،‬ﻧﻮﺩ ﺗﻮﺿﻴﺢ ﺑﻌﺾ ﺍﻷﻣﻮﺭ ﻣﻨﻬﺎ‪:‬‬
‫ﺍ‪ -‬ﻣﺸﻜﻞ ﻧﻘﺎﻁ ﺍﻻﻧﻄﻼﻕ‪ ،‬ﺍﻟﱵ ﳝﻜﻦ ﺣﻠﻬﺎ ﺑﺈﺣﺪﻯ ﺍﻻﻗﺘﺮﺍﺣﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪122‬‬
‫‪ (a‬ﻭﺿﻊ ﻛﻞ ﺍﻟﻘﻴ‪‬ﻢ ﺍﻻﺑﺘﺪﺍﺋﻴﺔ ﻣﺴﺎﻭﻳﺔ ﻟﻠﺼﻔﺮ‪ ،‬ﻭﻫﺬﺍ ﻳﻜﻮﻥ ﻣﻘﺒﻮﻻﹰ ﰲ ﺣﺎﻟﺔ‬
‫ﺗﻮﻓﺮ ﻛﻤﻴﺔ ﻣﻌﺘﱪﺓ ﻣﻦ ﺍﳌﺸﺎﻫﺪﺍﺕ‪.‬‬
‫ﻟﺬﻟﻚ‬ ‫‪p+1‬‬ ‫‪ (b‬ﺣﺴﺎﺏ ﺍﻟﻘﻴ‪‬ﻢ ﺍﻻﺑﺘﺪﺍﺋﻴﺔ‪ :‬ﺗﺒﺪﺃ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻤﻬﻴﺪ ﻣﻦ ﺍﻟﻔﺘﺮﺓ‬
‫ﻭﺟﺐ ﺗﻮﻓﲑ ﺍﻟﻘﻴ‪‬ﻢ ﺍﻻﺑﺘﺪﺍﺋﻴﺔ ﻟﻠﻌﻤﻠﻴﺔ ﻭﻛﻤﺎ ﻳﻠﻲ ‪:‬‬
‫‪8‬‬

‫‪~y = y‬‬
‫‪p‬‬ ‫‪p‬‬

‫) ‪rp = ( y p - y p- 1‬‬
‫‪yt‬‬
‫= ‪St‬‬
‫‪y‬‬

‫ﻓﻘﻂ‪.‬‬ ‫‪t = 1, 2,..., p‬‬ ‫ﺣﻴﺚ‬

‫‪1 p‬‬
‫=‪y‬‬ ‫‪åyt‬‬ ‫ﺃﻳﻦ‬
‫‪p t =1‬‬
‫ﺏ‪ -‬ﺻﻴﻐﺔ ﻣﻌﺎﺩﻟﺔ ﺍﻟﺘﻨﺒﺆ‪ ،‬ﳚﺐ ﺍﻹﺷﺎﺭﺓ ﻫﻨﺎ ﺇﱃ ﺿﺮﻭﺭﺓ ﺍﻻﺣﺘﻔﺎﻅ‬
‫ﺑﺎﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ ﺍﻷﺧﲑﺓ ﻻﺳﺘﻌﻤﺎﳍﺎ ﰲ ﺍﻟﺘﻨﺒﺆ ﺍﳌﺴﺘﻘﺒﻠﻲ‪ .‬ﺗﻜﻮﻥ ﻣﻌﺎﺩﻟﺔ ﺍﻟﺘﻨﺒﺆ‬
‫ﻛﺎﻵﰐ ﰲ ﺍﳊﺎﻟﺘﲔ ﺍﻟﺘﺠﻤﻴﻌﻴﺔ ﻭ ﺍﳉﺪﺍﺋﻴﺔ ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪:‬‬
‫~‬
‫‪yˆ T + L = ( y T + LrT ) + S (T + L ) - p‬‬ ‫)‪(4.31‬‬

‫~‬
‫‪yˆ T + L = ( y T + LrT ) S (T + L ) - p‬‬ ‫)‪(4.32‬‬

‫ﻭﰲ ﺣﺎﻟﺔ ) ‪ ( L = 1‬ﻓﺈﻥ ﺍﳌﻌﺎﺩﻟﺘﲔ ﺗﺼﺒﺤﺎﻥ ﻫﻜﺬﺍ‪:‬‬

‫~‬
‫‪yˆ T +1 = ( y T + rT ) + S ( T +1) - p‬‬
‫~‬
‫‪yˆ T +1 = ( y T + rT ) S ( T +1) - p‬‬

‫‪ - 8‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ‪:‬‬
‫‪a- Forecasting Economic Time Series, C. Granger and P. Newbold ,P. 166‬‬
‫‪b- Les Méthodes de Prevision en Economie ,Guy Aansion ,P. 194‬‬

‫‪123‬‬
‫ﺍﳉﺪﻭﻝ)‪(10‬‬ ‫ﺍﳉﺪﻭﻝ )‪ (30‬ﳝﺜﻞ ﺍﳊﻞ ﺍﻟﺬﻱ ﻳﻌﻄﻴﻪ ﻫﺬﺍ ﺍﻟﱪﻧﺎﻣﺞ ﳌﻌﻄﻴﺎﺕ‬
‫ﺻﻔﺤﺔ )‪ (46‬ﺍﳌﺘﻌﻠﻘﺔ ﺑﺎﳌﺘﻐﲑ ﺍﻻﺟﺘﻤﺎﻋﻲ ﺃﻳﻦ‪:‬‬
‫‪RSS = 366. 9‬‬ ‫ﻭ‬ ‫‪a = 0.89 , g = .38 , b = 0.9‬‬

‫ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ )ﺍﳊﺎﻟﺔ ﺍﳉﺪﺍﺋﻴﺔ(‪:‬‬

‫~‬
‫‪yˆ T +1 = ( y T + rT ) S (T +1) - p‬‬
‫~‬
‫‪yˆ T +1 = yˆ 21 = ( y 20 + r20 ) S17‬‬ ‫ﻭﻣﻨﻪ‬
‫‪= (53.96 + 11.86)0.38‬‬

‫‪= 25. 01‬‬

‫~‬
‫‪yˆ T + 2 = yˆ 22 = ( y 20 + 2{r20 }) S18‬‬

‫‪= (53.96 + 23.72)0.94‬‬ ‫ﻭ‬

‫‪= 73.02‬‬

‫ﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ ﳝﻜﻦ ﺣﺴﺎﺏ ﺍﻟﺘﻨﺒﺆﺍﺕ ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﺍﻹﺿﺎﻓﻴﺔ ﺍﻷﺧﺮﻯ‪.‬‬

‫ﺍﳉﺪﻭﻝ )‪ :(30‬ﻧﺘﺎﺋﺞ ﺍﻟﺘﻤﻬﻴﺪ‬
‫‪$y t‬‬ ‫‪St‬‬ ‫‪rt‬‬ ‫‪%y t‬‬ ‫ﺍﳌﺸﺎﻫﺪﺓ‬ ‫‪T‬‬
‫‪--‬‬ ‫‪0.656‬‬ ‫‪--‬‬ ‫‪--‬‬ ‫‪14‬‬ ‫‪1‬‬
‫‪--‬‬ ‫‪0.808‬‬ ‫‪--‬‬ ‫‪--‬‬ ‫‪20‬‬ ‫‪2‬‬
‫‪--‬‬ ‫‪1.777‬‬ ‫‪--‬‬ ‫‪--‬‬ ‫‪44‬‬ ‫‪3‬‬
‫‪--‬‬ ‫‪0.848‬‬ ‫‪--‬‬ ‫‪21‬‬ ‫‪21‬‬ ‫‪4‬‬
‫‪0.46-‬‬ ‫‪0.636‬‬ ‫‪16.34-‬‬ ‫‪15.51‬‬ ‫‪10‬‬ ‫‪5‬‬
‫‪10.28‬‬ ‫‪0.901‬‬ ‫‪8.11-‬‬ ‫‪20.83‬‬ ‫‪19‬‬ ‫‪6‬‬
‫‪.....‬‬ ‫‪.....‬‬ ‫‪.....‬‬ ‫‪.....‬‬ ‫‪.....‬‬ ‫‪.....‬‬
‫‪5.21‬‬ ‫‪0.38‬‬ ‫‪5.44-‬‬ ‫‪16.24‬‬ ‫‪6‬‬ ‫‪17‬‬
‫‪6.09‬‬ ‫‪0.94‬‬ ‫‪5.13-‬‬ ‫‪11.63‬‬ ‫‪11‬‬ ‫‪18‬‬
‫‪64.37‬‬ ‫‪2.333‬‬ ‫‪2.73‬‬ ‫‪27.19‬‬ ‫‪64‬‬ ‫‪19‬‬
‫‪57.81‬‬ ‫‪0.921‬‬ ‫‪11.86‬‬ ‫‪53.96‬‬ ‫‪50‬‬ ‫‪20‬‬

‫‪124‬‬
 ‫‪1‬‬ ‫ﲤﺮﻳﻦ‬
‫ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﻔﺼﻠﻴﺔ ]‪ [quarterly‬ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪w‬‬ ‫اﻟﺴﻨﺔ‬ ‫‪w‬‬ ‫اﻟﺴﻨﺔ‬ ‫‪w‬‬ ‫اﻟﺴﻨﺔ‬
‫‪32‬‬ ‫‪2004.1‬‬ ‫‪12‬‬ ‫‪1999.1‬‬ ‫‪9‬‬ ‫‪1994.1‬‬
‫‪39‬‬ ‫‪2004.2‬‬ ‫‪19‬‬ ‫‪1999.2‬‬ ‫‪20‬‬ ‫‪1994.2‬‬
‫‪90‬‬ ‫‪2004.3‬‬ ‫‪52‬‬ ‫‪1999.3‬‬ ‫‪54‬‬ ‫‪1994.3‬‬
‫‪21‬‬ ‫‪2004.4‬‬ ‫‪26‬‬ ‫‪1999.4‬‬ ‫‪30‬‬ ‫‪1994.4‬‬
‫‪23‬‬ ‫‪2005.1‬‬ ‫‪15‬‬ ‫‪2000.1‬‬ ‫‪4‬‬ ‫‪1995.1‬‬
‫‪42‬‬ ‫‪2005.2‬‬ ‫‪28‬‬ ‫‪2000.2‬‬ ‫‪14‬‬ ‫‪1995.2‬‬
‫‪77‬‬ ‫‪2005.3‬‬ ‫‪59‬‬ ‫‪2000.3‬‬ ‫‪41‬‬ ‫‪1995.3‬‬
‫‪30‬‬ ‫‪2005.4‬‬ ‫‪24‬‬ ‫‪2000.4‬‬ ‫‪22‬‬ ‫‪1995.4‬‬
‫‪21‬‬ ‫‪2006.1‬‬ ‫‪16‬‬ ‫‪2001.1‬‬ ‫‪12‬‬ ‫‪1996.1‬‬
‫‪44‬‬ ‫‪2006.2‬‬ ‫‪28‬‬ ‫‪2001.2‬‬ ‫‪11‬‬ ‫‪1996.2‬‬
‫‪112‬‬ ‫‪2006.3‬‬ ‫‪73‬‬ ‫‪2001.3‬‬ ‫‪46‬‬ ‫‪1996.3‬‬
‫‪31‬‬ ‫‪2006.4‬‬ ‫‪18‬‬ ‫‪2001.4‬‬ ‫‪21‬‬ ‫‪1996.4‬‬
‫‪36‬‬ ‫‪2007.1‬‬ ‫‪14‬‬ ‫‪2002.1‬‬ ‫‪12‬‬ ‫‪1997.1‬‬
‫‪60‬‬ ‫‪2007.2‬‬ ‫‪21‬‬ ‫‪2002.2‬‬ ‫‪14‬‬ ‫‪1997.2‬‬
‫‪80‬‬ ‫‪2007.3‬‬ ‫‪76‬‬ ‫‪2002.3‬‬ ‫‪34‬‬ ‫‪1997.3‬‬
‫‪47‬‬ ‫‪2007.4‬‬ ‫‪22‬‬ ‫‪2002.4‬‬ ‫‪25‬‬ ‫‪1997.4‬‬
‫‪43‬‬ ‫‪2008.1‬‬ ‫‪24‬‬ ‫‪2003.1‬‬ ‫‪10‬‬ ‫‪1998.1‬‬
‫‪77‬‬ ‫‪2008.2‬‬ ‫‪26‬‬ ‫‪2003.2‬‬ ‫‪28‬‬ ‫‪1998.2‬‬
‫‪70‬‬ ‫‪2008.3‬‬ ‫‪83‬‬ ‫‪2003.3‬‬ ‫‪51‬‬ ‫‪1998.3‬‬
‫‪49‬‬ ‫‪2008.4‬‬ ‫‪27‬‬ ‫‪2003.4‬‬ ‫‪19‬‬ ‫‪1998.4‬‬

‫ﺑﻴﺎﻧﻴﺎﹰ ﻭﺫﻟﻚ ﺑﺎﻻﺳﺘﻌﺎﻧﺔ‬ ‫)‪(W‬‬ ‫‪ -‬ﺍﳌﻄﻠﻮﺏ ﻛﺸﻒ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬

‫ﺑﱪﻧﺎﻣﺞ ‪ Eviews‬ﻭﺣﺴﺐ ﺍﻟﻠﻮﺣﺎﺕ ﺍﳌﺴﺎﻋﺪﺓ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ -(a‬ﺑﻌﺪ ﻓﺘﺢ ﺃﻳﻘﻮﻧﺔ ‪ ،Eviews‬ﻳﺘﻢ ﻣﺒﺎﺷﺮﺓ ﻓﺘﺢ ﻣﻠﻒ ﺟﺪﻳﺪ‬
‫)‪ workfile (New‬ﰒ ﲢﺪﻳﺪ ﺍﻟﻌﻴﻨﺔ ﻋﻦ ﻃﺮﻳﻖ ﺗﻌﻴﲔ ﻓﺘﺮﺓ ﺑﺪﺍﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ‬

‫‪125‬‬
‫ﻣﻊ ﲢﺪﻳﺪ ﻃﺒﻴﻌﺔ ﺍﳌﻌﻄﻴﺎﺕ‬ ‫)‪(Ending date‬‬ ‫ﻭ‪‬ﺎﻳﺘﻬﺎ‬ ‫ﺍﻟﺰﻣﻨﻴﺔ)‪(Starting date‬‬
‫ﺍﻟﻔﺼﻠﻴﺔ)‪. (Quarterly‬‬

‫‪126‬‬
‫‪ -(b‬ﺗﺴﻤﻴﺔ ﺍﳌﺘﻐﲑ ﻋﻦ ﻃﺮﻳﻖ ﺍﻟﺘﻌﻠﻴﻤﺔ )‪ (Data‬ﻭﺇﺩﺧﺎﻝ ﺍﳌﻌﻄﻴﺎﺕ ﰲ‬
‫ﺍﳋﺎﻧﺎﺕ ﺍﶈﺪﺩﺓ‪ .‬ﳝﻜﻦ ﻟﻠﻄﺎﻟﺐ ﺃﻥ ﻳﺴﻤﻲ ﺍﳌﺘﻐﲑ ﺍﳌﺪﺭﻭﺱ ‪ W‬ﺃﻭﻗﺪ ﳜﺘﺎﺭ ﺃﻱ‬
‫ﺗﺴﻤﻴﺔ ﺃﺧﺮﻯ ﺗﻨﺎﺳﺒﻪ‪.‬‬

‫‪ -(c‬ﺑﻌﺪ ﺍﻻﻧﺘﻬﺎﺀ ﻣﻦ ﺇﺩﺧﺎﻝ ﺍﳌﻌﻄﻴﺎﺕ‪ ،‬ﻳﺘﻢ ﺭﺳﻢ ﺍﳌﻨﺤﲎ ﺍﻟﺒﻴﺎﱐ‬

‫ﺑﺎﻻﺳﺘﻌﺎﻧﺔ ﺑﺎﻟﺘﻌﻠﻴﻤﺔ ‪ Graph‬ﰒ ﺍﺧﺘﻴﺎﺭ ﺷﻜﻞ ﺍﳌﻨﺤﲎ )ﰲ ﻫﺬﻩ ﺍﳋﺎﻟﺔ ‪ (Line‬ﻛﻤﺎ‬
‫ﺗﺒﻴﻨﻪ ﺍﻟﻠﻮﺣﺔ ﺍﻟﺘﺎﻟﻴﺔ‪.‬‬

‫‪127‬‬
‫ﺇﻥ ﺍﻟﺸﻜﻞ ﺃﻋﻼﻩ‪ ،‬ﻳﻌﻜﺲ ﺗﻄﺒﻴﻖ ﺍﻟﺘﻌﻠﻴﻤﺘﲔ ﺍﳌﺬﻛﻮﺭﺗﲔ ﺳﺎﻟﻔﺎﹰ‪ .‬ﻛﻤﺎ‬
‫ﻧﺴﺘﻨﺘﺞ ﻣﻦ ﺍﻟﺸﻜﻞ ﻭﺟﻮﺩ ﻣﺮﻛﺒﱵ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﺍﻟﻔﺼﻠﻴﺔ ﺇﺿﺎﻓﺔ ﺇﱃ ﺍﳌﺮﻛﺒﺔ‬
‫ﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪.‬‬

‫‪128‬‬
 ‫ﲤﺮﻳﻦ@‪2‬‬
‫‪ -‬ﻟﺪﻳﻚ ﺍﳌﻌﻄﻴﺎﺕ ‪ Vt‬ﺍﳋﺎﺻﺔ ﲟﺘﻐﲑ ﲡﺎﺭﻱ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪-1992.1‬‬
‫‪.1994.4‬‬
‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪t‬‬
‫‪184‬‬ ‫‪149‬‬ ‫‪138‬‬ ‫‪129‬‬ ‫‪157‬‬ ‫‪118‬‬ ‫‪Vt‬‬
‫‪12‬‬ ‫‪11‬‬ ‫‪10‬‬ ‫‪9‬‬ ‫‪8‬‬ ‫‪7‬‬ ‫‪t‬‬
‫‪190‬‬ ‫‪175‬‬ ‫‪218‬‬ ‫‪159‬‬ ‫‪160‬‬ ‫‪155‬‬ ‫‪Vt‬‬
‫ﺍﻟﻤﻁﻠﻭﺏ‪:‬‬
‫ﺍ ‪-‬ﺍﺴﺘﻨﺘﺎﺝ ﻤﺭﻜﺒﺎﺕ ﻫﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﺒﻴﺎﻨﻴﺎ‪.‬‬

‫ﺏ ‪ -‬ﺗﺪﻋﻴﻢ ﺍﻟﻨﺘﻴﺠﺔ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﺧﺘﺒﺎﺭﻱ ﺩﺍﻧﻴﺎﻝ ﻭﻛﺮﻳﺴﺘﺎﻝ‪ -‬ﻭﻟﻴﺲ ﻛﻞ‬

‫ﰲ ﳎﺎﻟﻪ‪.‬‬
‫ﺟـ ‪ -‬ﺣﺴﺎﺏ ﺍﳌﺆﺷﺮﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻄﺮﻳﻘﺔ ﺍﻟﻨﺴﺐ ﺍﳌﻮﲰﻴﺔ‪ ،‬ﰒ ﺇﳚﺎﺩ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻨﻬﺎ‪.‬‬
‫ﺩ ‪ -‬ﺑﻌﺪ ﺗﻘﺪﻳﺮ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻨﺎﲡﺔ ﺃﻋﻼﻩ ﺏ ‪ OLS‬ﰎ ﺍﻟﺘﻮﺻﻞ ﺇﱃ ﺍﻟﻨﺘﺎﺋﺞ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪V at = 124. 86 + 5. 56 t‬‬
‫) ‪( 4. 99‬‬ ‫) ‪( 0. 68‬‬
‫‪R 2 = 0. 87 f - stat = 67. 22‬‬

‫) ‪( V T+1‬‬ ‫ﻗﻴ‪‬ﻢ ﺍﻟﻨﻤﻮﺫﺝ ﺇﺣﺼﺎﺋﻴﺎ‪ ،‬ﻭﺇﻥ ﻛﺎﻥ ﻣﻮﻓﻘﺎ ﻓﺎﺳﺘﻌﻤﻠﻪ ﻟﻠﺘﻨﺒﺆ ﺑـ‬

‫‪129‬‬
 ‫ﲤﺮﻳﻦ@‪3‬‬
‫‪ -‬ﺑﺎﺳﺘﺨﺪﺍﻡ ﻧﻔﺲ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﻟﻮﺍﺭﺩﺓ ﰲ ﺍﻟﺴﺆﺍﻝ ﺍﻷﻭﻝ ﺃﺫﻛﺮ ﺍﻟﻄﺮﻳﻘﺔ‬
‫ﺍﳌﻨﺎﺳﺒﺔ ﻭﺍﳌﺒﺎﺷﺮﺓ ﻟﻠﺘﻤﻬﻴﺪ ﻭﺍﻟﺘﻨﺒﺆ ﺑﺎﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ‪ ،‬ﻣﻊ ﻛﺘﺎﺑﺔ ﺍﳌﻌﺎﺩﻻﺕ‬
‫ﺍﻟﻀﺮﻭﺭﻳﺔ‪ ،‬ﰒ ﺑﲔ ﻃﺮﻳﻘﺔ ﺣﺴﺎﺏ ﻗﻴ‪‬ﻢ ﺍﻻﻧﻄﻼﻕ ﺇﻥ ﻛﺎﻧﺖ ‪. P=4‬‬

‫‪ -‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ ﺣﻮﻝ ﺃﺳﻌﺎﺭ ﺍﻟﱪﺍﻧﺖ )‪ (Brent Prices‬ﰲ‬

‫ﺃﺣﺪ ﺍﻷﺳﻮﺍﻕ ﺍﻟﺪﻭﻟﻴﺔ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪:1994.12-1993.01‬‬

‫‪yt‬‬ ‫‪t‬‬ ‫‪yt‬‬ ‫‪t‬‬ ‫‪yt‬‬ ‫‪t‬‬ ‫‪yt‬‬ ‫‪t‬‬

‫‪22.27‬‬ ‫‪19‬‬ ‫‪17.61‬‬ ‫‪13‬‬ ‫‪16.76‬‬ ‫‪7‬‬ ‫‪17.34‬‬ ‫‪1‬‬
‫‪24.08‬‬ ‫‪20‬‬ ‫‪18.52‬‬ ‫‪14‬‬ ‫‪16.71‬‬ ‫‪8‬‬ ‫‪18.45‬‬ ‫‪2‬‬
‫‪20.59‬‬ ‫‪21‬‬ ‫‪18.01‬‬ ‫‪15‬‬ ‫‪16.03‬‬ ‫‪9‬‬ ‫‪18.73‬‬ ‫‪3‬‬
‫‪20.73‬‬ ‫‪22‬‬ ‫‪20.17‬‬ ‫‪16‬‬ ‫‪16.53‬‬ ‫‪10‬‬ ‫‪18.64‬‬ ‫‪4‬‬
‫‪20.87‬‬ ‫‪23‬‬ ‫‪21.58‬‬ ‫‪17‬‬ ‫‪15.10‬‬ ‫‪11‬‬ ‫‪18.46‬‬ ‫‪5‬‬
‫‪19.25‬‬ ‫‪24‬‬ ‫‪21.77‬‬ ‫‪18‬‬ ‫‪13.54‬‬ ‫‪12‬‬ ‫‪17.74‬‬ ‫‪6‬‬

‫ﺍﳌﻄﻠﻮﺏ‪ :‬ﺇﺫﺍ ﺗﻮﻓﺮﺕ ﻟﺪﻳﻚ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ‪ ،‬ﺗﻨﺒﺄ ﺑﺄﺳﻌﺎﺭ ﻫﺬﻩ ﺍﳌﺎﺩﺓ‬
‫ﻟﻠﺸﻬﺮﻳﻦ ﺍﻟﺘﺎﻟﻴﲔ ﻣﺒﻴﻨﺎﹰ ﻛﻞ ﺍﳌﺮﺍﺣﻞ ﺍﻟﱵ ﲤﺮ ‪‬ﺎ‪.‬‬
‫~‬ ‫~‬
‫‪y 23 = 19 . 65 , y 24 = 21 . 22 ,‬‬

‫‪a = 0. 94, b = 0. 2, g = 0. 03‬‬

‫‪r23 = 0. 344‬‬
‫‪S 12 = 0. 97, S 8 = 0. 98,‬‬

‫‪130‬‬
 ‫∑‪4@ÂÌä‬‬
‫ﰲ ﺃﺣﺪ ﺍﻷﺳﻮﺍﻕ‬ ‫‪butane‬‬ ‫‪ -‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ ﺣﻮﻝ ﺃﺳﻌﺎﺭ‬
‫ﺍﻟﺪﻭﻟﻴﺔ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪:1994.12-1992.01‬‬

‫‪151‬‬ ‫‪290‬‬ ‫‪275.83‬‬ ‫‪207.60‬‬ ‫‪21.09‬‬ ‫‪217.74‬‬ ‫‪1992.01‬‬

‫‪157.5‬‬ ‫‪147.5‬‬ ‫‪159.90‬‬ ‫‪207.60‬‬ ‫‪90.37‬‬ ‫‪121.97‬‬ ‫‪1992.07‬‬
‫‪160‬‬ ‫‪183‬‬ ‫‪196.80‬‬ ‫‪19.88‬‬ ‫‪105.74‬‬ ‫‪142.50‬‬ ‫‪1993.01‬‬
‫‪152‬‬ ‫‪168‬‬ ‫‪171.75‬‬ ‫‪20.30‬‬ ‫‪103.40‬‬ ‫‪130.50‬‬ ‫‪1993.07‬‬
‫‪131‬‬ ‫‪133‬‬ ‫‪134.75‬‬ ‫‪18.73‬‬ ‫‪100.11‬‬ ‫‪128.50‬‬ ‫‪1994.01‬‬
‫‪158‬‬ ‫‪130‬‬ ‫‪121.38‬‬ ‫‪16.71‬‬ ‫‪71.29‬‬ ‫‪116.50‬‬ ‫‪1994.07‬‬

‫ﺍﳌﻄﻠﻮﺏ‪ :‬ﺗﻨﺒﺄ ﺑﺄﺳﻌﺎﺭ ﻫﺬﻩ ﺍﳌﺎﺩﺓ ﻟﻠﺸﻬﺮﻳﻦ ﺍﻟﺘﺎﻟﻴﲔ ﻣﺒﻴﻨﺎﹰ ﳐﺘﻠﻒ ﻣﺮﺍﺣﻠﻪ‪.‬‬
‫‪-‬‬
‫‪s i = 1.2 + 0.03 y i‬‬
‫~‬
‫‪y 36 = 139.65,‬‬
‫‪a = 0.94, b = 0.2, g = 0.03‬‬
‫‪r36 = 0.344‬‬
‫‪S 25 = 21.64, S 26 = -63, S 35 = 28.64, S 36 = 35.57‬‬

‫ﲤﺮﻳﻦ@‪5‬‬
‫‪ -‬ﺑﺮﺭ ﺍﺳﺘﻌﻤﺎﻝ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﻨﺒﺆﻳﺔ ﺍﻟﺘﺎﻟﻴﺔ ﰲ ﳕﻮﺫﺝ ﻫﻮﻟﺖ ‪:‬‬
‫~‬ ‫~‬
‫‪y T + L = y T + L . rT‬‬

‫ﺣﻴﺚ ‪ L‬ﳝﺜﻞ ﺃﻓﻖ ﺍﻟﺘﻨﺒﺆ‪ .‬ﰒ ﺃﺫﻛﺮ ﻋﻼﻗﺔ ﻫﺬﺍ ﺍﻷﺧﲑ ﺑﻨﻤﻮﺫﺝ ﺍﻟﺘﻨﺒﺆ ﰲ‬
‫ﺣﺎﻟﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﶈﻠﻲ‪.‬‬

‫‪131‬‬
 ‫ﲤﺮﻳﻦ@‪6‬‬

‫ﻣﺴﺘﻌﻴﻨﺎ ﺑﱪﻧﺎﻣﺞ ‪ ،Eviews‬ﻭﻣﺴﺘﺨﺪﻣﺎ ﻣﻌﻄﻴﺎﺕ ﺇﻧﺘﺎﺝ ﺍﳊﺒﻮﺏ)ﺃﻟﻒ‬

‫ﻃﻦ( ﰲ ﺍﳉﺰﺍﺋﺮ ﻣﻦ ‪ 1963‬ﺇﱃ ‪ ،2006‬ﺍﳌﻄﻠﻮﺏ‪:‬‬
‫ﺃ‪ -‬ﻛﺸﻒ ﻣﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ؛‬
‫ﺏ‪ -‬ﲤﻬﻴﺪ ﺍﻟﺴﻠﺴﻠﺔ ﺑﺘﻘﻨﻴﺔ ﻣﻨﺎﺳﺒﺔ؛‬
‫‪.2009 ،2008 ،2007‬‬ ‫ﺕ‪ -‬ﺑﻌﺪ ﺗﻘﻴﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ‪ ،‬ﺍﻟﺘﻨﺒﺆ ﻟﺴﻨﻮﺍﺕ‬
‫ﺙ‪ -‬ﺣﺴﺎﺏ ﺧﻄﺄ ﺍﻟﺘﻨﺒﺆ ﺇﺫﺍ ﺑﻠﻎ ﺳﻨﺔ ‪ 2009‬ﺍﻹﻧﺘﺎﺝ ‪ 6.1‬ﻣﻠﻴﻮﻥ ﻃﻦ‪.‬‬
‫إﻧﺗﺎج‬
‫اﻟﺳﻧﺔ‬ ‫إﻧﺗﺎج اﻟﺣﺑوب‬ ‫اﻟﺳﻧﺔ‬ ‫إﻧﺗﺎج اﻟﺣﺑوب‬ ‫اﻟﺳﻧﺔ‬ ‫إﻧﺗﺎج اﻟﺣﺑو ب‬ ‫اﻟﺳﻧﺔ‬
‫اﻟﺣﺑوب‬
‫‪4901‬‬ ‫‪1996‬‬ ‫‪1460‬‬ ‫‪1985‬‬ ‫‪1595‬‬ ‫‪1974‬‬ ‫‪2360‬‬ ‫‪1963‬‬
‫‪870‬‬ ‫‪1997‬‬ ‫‪2917‬‬ ‫‪1986‬‬ ‫‪1480‬‬ ‫‪1975‬‬ ‫‪2320‬‬ ‫‪1964‬‬
‫‪3025‬‬ ‫‪1998‬‬ ‫‪2402‬‬ ‫‪1987‬‬ ‫‪2680‬‬ ‫‪1976‬‬ ‫‪1488‬‬ ‫‪1965‬‬
‫‪2021‬‬ ‫‪1999‬‬ ‫‪2065‬‬ ‫‪1988‬‬ ‫‪2313‬‬ ‫‪1977‬‬ ‫‪1734‬‬ ‫‪1966‬‬
‫‪934‬‬ ‫‪2000‬‬ ‫‪2003‬‬ ‫‪1989‬‬ ‫‪1442‬‬ ‫‪1978‬‬ ‫‪776‬‬ ‫‪1967‬‬
‫‪2659‬‬ ‫‪2001‬‬ ‫‪1625‬‬ ‫‪1990‬‬ ‫‪1538‬‬ ‫‪1979‬‬ ‫‪1641‬‬ ‫‪1968‬‬
‫‪1953‬‬ ‫‪2002‬‬ ‫‪3808‬‬ ‫‪1991‬‬ ‫‪1619‬‬ ‫‪1980‬‬ ‫‪2128‬‬ ‫‪1969‬‬
‫‪4266‬‬ ‫‪2003‬‬ ‫‪3329‬‬ ‫‪1992‬‬ ‫‪2421‬‬ ‫‪1981‬‬ ‫‪1852‬‬ ‫‪1970‬‬
‫‪4033‬‬ ‫‪2004‬‬ ‫‪1452‬‬ ‫‪1993‬‬ ‫‪1832‬‬ ‫‪1982‬‬ ‫‪2058‬‬ ‫‪1971‬‬
‫‪3527‬‬ ‫‪2005‬‬ ‫‪963‬‬ ‫‪1994‬‬ ‫‪1523‬‬ ‫‪1983‬‬ ‫‪1735‬‬ ‫‪1972‬‬
‫‪4018‬‬ ‫‪2006‬‬ ‫‪2138‬‬ ‫‪1995‬‬ ‫‪1298‬‬ ‫‪1984‬‬ ‫‪2362‬‬ ‫‪1973‬‬

‫‪132‬‬
 ‫اﻟﺒﺎب اﻟﺜﺎﻧﻲ‬
‫ﲢﻠﻴﻞ‬
‫اﻟﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻌﺸﻮاﺋﻴﺔ‬
134
 ‫ﺍﻟﻔﺼﻞ ﺍﳋﺎﻣﺲ‬

‫ﻃﺮﻳﻘﺔ ﺑﻮﻛﺲ ‪ -‬ﺟﺎﻧﻜﻴﻨﺰ ﻭﺍﻹﺳﺘﻘﺮﺍﺭﻳﺔ‬

‫ﺫﻛﺮﻧﺎ ﺿﻤﻦ ﺍﻟﻔﺼﻞ ﺍﻷﻭﻝ ﳍﺬﺍ ﺍﻟﻜﺘﺎﺏ ﻋﻨﺪ ﺗﻨﺎﻭﻝ ﻣﻮﺿﻮﻉ ﺃﻧﻮﺍﻉ‬
‫ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺘﻨﺒﺆﻳﺔ‪ ،‬ﺃﻧﻪ ﰲ ﳕﺎﺫﺝ ﺍﻟﺴﻼﺳــﻞ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﻋﻨﺪ ﺩﺭﺍﺳﺔ ﺍﻟﻈﺎﻫﺮﺓ‬
‫ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻳﺘﻢ ﺍﻻﻋﺘﻤﺎﺩ ﻋﻠﻰ ﲢﻠﻴﻞ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻟﻔﻬﻢ ﺳﻠﻮﻛﻬﺎ ﺍﳌﺎﺿﻲ ﰒﹶ ﺑﻨﺎﺀ‬
‫ﳕﻮﺫﺝ ﺧﺎﺹ ﺑﻌﻤﻠﻴﺎﺕ ﺍﻟﺘﻨﺒﺆ‪.‬‬

‫ﻛﻮ‪‬ﺎ ﻛﺬﻟﻚ )ﺃﻱ ﺍﻟﻨﻤﺎﺫﺝ(‪ ،‬ﻭﺭﻏﻢ ﺃ‪‬ﺎ ﻻ ﺗﺴﺘﻌﲔ ﺑﻜﻞ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﻟﱵ‬
‫ﺗﺴﺘﺨﺪﻣﻬﺎ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ‪ -‬ﻣﻦ ﻣﻌﻄﻴﺎﺕ ﺣﻮﻝ ﺍﳌﺘﻐﲑ ﺍﳌﺮﺍﺩ ﺩﺭﺍﺳﺘﻪ ﻭﻛﺬﺍ‬
‫ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﳋﺎﺻﺔ ﻭﺍﶈﻴﻄﺔ ﺑﺎﻟﻈﺎﻫﺮﺓ‪ ،‬ﺍﻟﱵ ﻧﺴﻤﻴﻬﺎ ﺑﺎﳌﺘﻐﲑﺍﺕ ﺍﳌﺴﺘﻘﻠﺔ ﻭﺍﳌﺆﺛﺮﺓ ﰲ‬
‫ﺍﳌﺘﻐﲑ ﺍﻟﺘﺎﺑﻊ )ﺍﻟﻈﺎﻫﺮﺓ(‪ -‬ﺇﻻ ﺃ‪‬ﺎ ﺃﺛﺒﺘﺖ ﺟﺪﺍﺭ‪‬ﺎ ﰲ ﺍﳌﻴﺪﺍﻥ ﺍﻟﺘﻨﺒﺌﻲ ﺍﻟﻘﺼﲑ ﺍﳌﺪﻯ‪.‬‬

‫ﺇﻥ ﺍﻟﺘﻔﺮﻗﺔ ﺍﻵﻥ ﺑﲔ ﳕﺎﺫﺝ ﺍﻻﺳﺘﻘﻄﺎﺏ ﻭﳕﺎﺫﺝ ﺑﻮﻛﺲ‪-‬ﺟﺎﻧﻜﻴﻨﺰ‪،‬‬

‫ﻛﺎﻟﺘﻔﺮﻗﺔ ﺑﲔ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﳊﺮﺓ )‪ (Non parametric‬ﻭﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﻏﲑ ﺍﳊﺮﺓ‬
‫)‪ .(Parametric‬ﳍﺬﺍ ﻓﺈﻥ ﻫﺬﻩ ﺍ‪‬ﻤﻮﻋﺔ ﺍﻷﺧﲑﺓ ﺗﻌﺘﱪ ﻣﻨﺎﻓﺲ ﺣﻘﻴﻘﻲ ﻟﻨﻤﺎﺫﺝ‬
‫ﺍﻻﺳﺘﻘﻄﺎﺏ‪ ،‬ﺭﻏﻢ ﺻﻌﻮﺑﺔ ﺍﻟﺘﻌﺎﻣﻞ ﻣﻌﻬﺎ ﻣﻦ ﺣﻴﺚ ﺍﻟﺘﺤﺪﻳﺪ‪ ،‬ﺍﻟﺘﻘﺪﻳﺮ ﻭﺍﻻﺧﺘﺒﺎﺭ‪،‬‬
‫ﻣﻘﺎﺭﻧﺔ ﺑﺎﻟﻨﻤﺎﺫﺝ ﺍﳌﻜﻴ‪‬ﻔﺔ ﺍﻟﻮﺍﺭﺩﺓ ﺿﻤﻦ ﳕﺎﺫﺝ ﺍﻻﺳﺘﻘﻄﺎﺏ ﺍﻟﱵ ﻻ ﺗﻮﻓﺮ ﻭﺑﺴﺒﺐ‬
‫ﻋﺪﻡ ﺧﻀﻮﻋﻬﺎ ﻟﻌﻤﻠﻴﺔ ﺍﻟﺘﻘﺪﻳﺮ‪ ،‬ﳎﺎﻻﺕ ﺍﻟﺜﻘﺔ‪ ،‬ﺍﻟﱵ ﺗﺴﺎﻫﻢ ﺑﺸﻜﻞ ﻛﺒﲑ ﰲ‬
‫ﺍﻻﺧﺘﻴﺎﺭ ﺍﳌﻮﻓﻖ ﻟﻠﻨﻤﻮﺫﺝ‪ .‬ﳍﺬﺍ ﻓﺈﻥ ﳕﺎﺫﺝ ﺍ‪‬ﻤﻮﻋﺔ ﺍﻟﺜﺎﻧﻴﺔ ﺍﳌﺴﻤﺎﺓ ﺑﺒﻮﻛﺲ‪-‬‬
‫ﺟﺎﻧﻜﻴﻨﺰ )‪ (Box-Jenkins‬ﲢﺘﺎﺝ ﺇﱃ ﺇﻣﻜﺎﻧﻴﺎﺕ ﻣﺎﺩﻳﺔ ﻭﺑﺸﺮﻳﺔ ﻣﺘﺨﺼﺼﺔ‪ ،‬ﺗﻘﻮﻡ‬
‫ﲟﻬﺎﻡ ﺍﻟﺘﻨﺒﺆ ﰲ ﺍﳌﺆﺳﺴﺎﺕ ﺍﳊﺪﻳﺜﺔ‪ ،‬ﺍﳌﺘﻮﺳﻄﺔ ﻭﺍﻟﻜﺒﲑﺓ‪.‬‬

‫‪135‬‬
‫ﻟﻘﺪ ﰎ ﰲ ﻫﺬﺍ ﺍﻟﺒﺎﺏ ﺃﺧﺬ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ ﻣﻮﺿﻮﻉ ﺍﻹﺳﺘﻘﺮﺍﺭﻳﺔ‬
‫)‪ (stationarity‬ﻛﻮﻥ ﺃﻏﻠﺒﻴﺔ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﻗﺪ ﺗﻮﻟﺪﺕ ﺃﻭ ﲤﺨﻀﺖ ﻣﻦ ﺑﻴﺌﺔ‬
‫ﻏﲑ ﻣﺴﺘﻘﺮﺓ‪ .‬ﻫﺬﻩ ﺍﻟﻈﺎﻫﺮﺓ ﻗﺪ ﻃﺮﺣﺖ ﺇﺷﻜﺎﻻﺕ ﻛﺜﲑﺓ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻌﻼﻗﺎﺕ‬
‫ﺍﻟﻮﳘﻴﺔ ﻭﻛﺬﺍ ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﺍﻟﺸﻌﺎﻋﻴﺔ‪.‬‬
‫‪ - 1‬ﺧﺼﺎﺋﺺ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪:‬‬
‫ﺇﻥ ﻋﻤﻠﻴﺔ ﺍﻟﺘﺤﻠﻴﻞ ﰲ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ‪ ،‬ﻛﻐﲑﻫﺎ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻷﺧﺮﻯ ‪‬ﺘﻢ‬
‫ﺑﺎﺳﺘﺨﻼﺹ ﺍﳋﺼﺎﺋﺺ ﺍﳉﻮﻫﺮﻳﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﺑﻐﻴﺔ ﺍﻻﺳﺘﻔﺎﺩﺓ ﻣﻨﻬﺎ ﻷﻏﺮﺍﺽ‬
‫ﺍﻟﻨﻤﺬﺟﺔ ﻓﻴﻤﺎ ﺑﻌﺪ‪ ،‬ﻭﻣﻦ ﻫﺬﻩ ﺍﳋﺼﺎﺋﺺ‪:‬‬

‫‪ -1‬ﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪ :‬ﺗﺘﻤﺜﻞ ﰲ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ‪ ،‬ﺍﻟﱵ ﳚﺐ ﺃﻥ ﺗﻜﻮﻥ ﻗﺪ ﺗﻮﻟﺪﺕ‬

‫ﻋﻦ ﻇﺮﻭﻑ ﻋﺸﻮﺍﺋﻴﺔ‪ .‬ﺇﺫﺍ ﻭﺑﺎﻓﺘﺮﺍﺽ ﺃﻧﻪ ﻟﺪﻳﻨﺎ ‪ -‬ﻭﻓﻖ ﻣﻨﻬﺠﻴﺔ ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ‪-‬‬
‫ﺳﻠﺴﻠﺔ ‪ y t‬ﺫﺍﺕ ﻣﺮﻛﺒﺘﲔ ﻋﺸﻮﺍﺋﻴﺔ ﻭﺍﲡﺎﻩ ﻋﺎﻡ‪ ،‬ﻭ ﺑﻌﺪ ﺃﺧﺬ ﻓﺮﻭﻗﻬﺎ ﻣﻦ ﺍﻟﺪﺭﺟﺔ‬
‫ﺍﻷﻭﱃ ﳓﺼﻞ ﻋﻠﻰ ﺳﻠﺴﻠﺔ ﻋﺸﻮﺍﺋﻴﺔ ﲝﺘﺔ ﻛﺎﻵﰐ ‪:‬‬

‫‪wt = e t‬‬

‫‪y t - y t -1 = e t‬‬

‫‪w t = y t - y t -1‬‬ ‫ﺣﻴﺚ‬

‫ﺣﻴﺚ ﳝﻜﻦ ﺍﻻﺳﺘﻌﺎﻧﺔ ﺑﺎﻟﻔﺮﺿﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫‪E( e t ) = 0‬‬ ‫)‪(5.2‬‬

‫‪E(e t , e s) = 0‬‬ ‫‪" t¹s‬‬ ‫)‪(5.3‬‬

‫‪2‬‬
‫‪= se‬‬ ‫‪" t =s‬‬ ‫)‪(5.4‬‬

‫‪136‬‬
‫ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻷﺧﲑ )‪ (5.1‬ﻳﺴﻤﻰ ﺑﻨﻤﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ‬
‫)‪ ،(Random Walk Process‬ﻛﻤﺎ ﻳﻄﻠﻖ ﻋﻠﻴﻪ ﳕﻮﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﻣﻦ ﺍﻟﺪﺭﺟﺔ‬
‫ﺍﻷﻭﱃ ﲟﻌﻠﻤﺔ ﺃﺣﺎﺩﻳﺔ )‪ AR(1‬ﺑﺘﻌﺒﲑ )‪.(Box-Jenkins‬‬
‫ﺑﺎﻟﺘﻌﻮﻳﺾ ﺍﳋﻠﻔﻲ ﺃﻭ ﺍﻟﺘﺮﺍﺟﻌﻲ )‪ (Back substitution‬ﰲ ﺍﻟﻨﻤﻮﺫﺝ )‪(5.1‬‬
‫ﳓﺼﻞ ﻋﻠﻰ‪:‬‬
‫‪y t = ( y t - 2 + e t -1 ) + e t‬‬

‫‪= y t - 2 + e t -1 + e t‬‬

‫‪y t = ( y t - 3 + e t - 2 ) + e t -1 + e t‬‬

‫‪= y t - 3 + e t - 2 + e t -1 + e t‬‬

‫‪j -1‬‬
‫‪yt = yt -j + å et -i‬‬ ‫)‪(5.5‬‬
‫‪i=0‬‬

‫ﺑﺎﻟﺮﺟﻮﻉ ﺍﳋﻠﻔﻲ ﺇﱃ ﺑﺪﺍﻳﺔ ﺗﻜﻮﻳﻦ ﻫﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ‪ ،‬ﻭﻓﻖ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﳝﻜﻦ‬
‫ﻛﺘﺎﺑﺔ )‪ (5.5‬ﰲ ﺍﻟﺸﻜﻞ ﺍﳉﺪﻳﺪ ﺍﻟﺘﺎﱄ ﳌﺎ ‪: t = j‬‬
‫‪t -1‬‬
‫‪y t = å e t - i + y0‬‬ ‫)‪(5.6‬‬
‫‪i=0‬‬

‫ﺣﻴﺚ ‪ y0‬ﻫﻲ ﻧﻘﻄﺔ ﺑﺪﺍﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ‪ ،‬ﻭﺍﳉﺰﺀ ﺍﻟﺜﺎﱐ ﻣﺎ ﻫﻮ ﺇﻻ ﳎﻤﻮﻉ‬

‫ﻋﻨﺎﺻﺮ ﻋﺸﻮﺍﺋﻴﺔ‪ .‬ﺇﺫﺍﹼ ﻭﺑﺈﺩﺧﺎﻝ ﺍﻟﺘﻮﻗﻊ )ﺍﻷﻣﻞ ﺍﻟﺮﻳﺎﺿﻲ( ﻋﻠﻰ ﺍﻟﻌﻼﻗﺔ )‪(5.6‬‬
‫ﳓﺼﻞ ﻋﻠﻰ‪:‬‬
‫‪E( yt ) = y = m‬‬ ‫)‪(5.7‬‬
‫‪0‬‬

‫ﺣﻴﺚ ‪m‬ﻫﻮ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ‪ ،‬ﺑﻴﻨﻤﺎ ﺗﺒﺎﻳﻨﻬﺎ ﻣﻌﻄﻰ ﺑـ‪:‬‬

‫‪t -1‬‬
‫‪var( y t ) = E ( y t - y 0 ) 2 = E [ å e t - i ] 2‬‬
‫‪i=0‬‬

‫‪137‬‬
 ‫ﳒﺪ ﺃﻥ‪:‬‬ ‫ﺍﻟﻔﺮﺿﻴﺔ )‪ (5.2‬ﻭ )‪(5.3‬‬ ‫ﺑﺎﺳﺘﻌﻤﺎﻝ‬
‫‪var ( y t ) = E ( e t + e t -1 + .. .. + e 1 ) 2‬‬

‫] ‪= [ s e2 + s e2 + ..... . + s e2‬‬

‫‪= ts‬‬ ‫‪2‬‬

‫‪e‬‬
‫)‪(5.8‬‬

‫ﲝﻜﻢ ﺃﻥ ﻫﺬﻩ ﺍﻟﻜﻤﻴﺔ ﻣﺮﺗﺒﻄﺔ ﺑﺎﳌﺘﻐﲑ )‪ (t‬ﺳﻴﻜﻮﻥ ﺍﻟﺘﺒﺎﻳﻦ ﻏﲑ ﺛﺎﺑﺖ ﺑﻞ‬

‫ﻣﺘﺰﺍﻳﺪﺓ ﻣﻊ ﺍﻟﺰﻣﻦ ﻭﻋﺎﻛﺴﺎﹰ ﻋﻨﺼﺮﺍ ﻣﻦ ﻋﻨﺎﺻﺮ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ‪.‬‬
‫ﻭﻓﻖ ﺍﻟﻨﻤﻮﺫﺝ )‪ ،(5.1‬ﳝﻜﻦ ﺇﳚﺎﺩ ﺗﻨﺒﺆ ﰲ ﺍﻟﻔﺘﺮﺓ ‪ T+1‬ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫ˆ‪y‬‬ ‫‪T +1‬‬
‫‪= E[ y‬‬
‫‪T +1‬‬
‫‪/‬‬ ‫‪y ,y‬‬
‫‪T‬‬ ‫‪T -1‬‬
‫‪,...,‬‬ ‫‪y‬‬ ‫‪1‬‬
‫]‬

‫ﲟﻌﻠﻮﻣﺔ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﻟﺘﺎﺭﳜﻴﺔ‬ ‫‪y T+1‬‬ ‫ﲟﻌﲎ ﺃﻥ ‪) $y T+1‬ﺍﻟﺘﻨﺒﺆ( ﻣﺎ ﻫﻮ ﺇﻻ ﺗﻮﻗﻊ‬

‫ﻟﻨﻔﺲ ﺍﳌﺘﻐﲑ‪ .‬ﺇﺫﺍ ﻭﺑﻌﺪ ﺍﻟﺘﻌﻮﻳﺾ ﰲ )‪:(5.1‬‬

‫ˆ‪y‬‬ ‫‪T +1‬‬

‫[‪= E‬‬ ‫‪y‬‬ ‫‪T‬‬
‫‪+‬‬ ‫‪e‬‬ ‫‪T +1‬‬
‫]‬

‫)‪(5.2‬‬ ‫ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻔﺮﺿﻴﺔ‬

‫ˆ‪y‬‬ ‫‪T +1‬‬
‫‪= yT‬‬

‫ˆ‪y‬‬ ‫‪T +2‬‬

‫[‪= E‬‬ ‫‪y‬‬ ‫‪T+2‬‬
‫‪/‬‬ ‫‪y ,y‬‬
‫‪T‬‬ ‫‪T -1‬‬
‫‪,...,‬‬ ‫‪y‬‬ ‫‪1‬‬
‫]‬ ‫ﻭ‬
‫] ‪= E [ y T+1 + e T+ 2‬‬

‫] ‪= E [ y T + e T+1 + e T+ 2‬‬

‫‪= yT‬‬
‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫ˆ‪y‬‬ ‫‪= YT‬‬ ‫)‪(5.9‬‬
‫‪T +L‬‬

‫‪138‬‬
‫ﺃﻱ ﺃﻥ ﺍﻟﺘﻨﺒﺆ ﻣﺎ ﻫﻮ ﺇﻻ ﺻﻮﺭﺓ ﻟﻠﺤﺎﺿﺮ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﻓﻬﻮ ﺗﻌﺒﲑ ﻋﻦ ﺃﺣﺪﺙ‬
‫ﻣﺸﺎﻫﺪﺓ ﻣﺘﻮﻓﺮﺓ ﻟﻼﺳﺘﺨﺪﺍﻡ‪ ،‬ﻭﻣﻨﻪ ﳝﻜﻦ ﺍﻟﻘﻮﻝ ﺃﻥ ﺍﻟﺘﻨﺒﺆ ﺍﳌﺴﺘﻘﺒﻠﻲ ﻟﻠﻤﺘﻐﲑ‬
‫ﺍﳌﺪﺭﻭﺱ ﺛﺎﺑﺖ‪ ،‬ﺭﻏﻢ ﺯﻳﺎﺩﺓ ﺗﺒﺎﻳﻨﻪ ﻣﻊ ﺍﻷﻓﻖ ‪.L‬‬
‫ﻣﻦ )‪ (5.8‬ﻧﺴﺘﻨﺘﺞ‪:‬‬
‫‪var( y t ) = t s e2‬‬
‫ﺃﻱ‬
‫‪var ( y T ) = T s e2‬‬

‫(‪var‬‬ ‫‪y‬‬ ‫‪) = (T + 1) s‬‬

‫‪T +1‬‬
‫‪2‬‬
‫‪e‬‬
‫(‪= var‬‬ ‫‪y‬‬ ‫‪T‬‬
‫)‪) + var( e1‬‬ ‫ﺇﺫﺍ‬
‫ﺣﻴﺚ ‪ e1‬ﳝﺜﻞ ﺧﻄﺄ ﺗﻨﺒﺆ ﺍﻟﻔﺘﺮﺓ ﺍﳌﻘﺒﻠﺔ‪ ،‬ﻭﺃﻳﻀﺎﹰ‬
‫‪var( y T + 2 ) = ( T + 2 ) s e2‬‬
‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫‪var( y T + L ) = ( T + L ) s e2‬‬

‫‪var( y T + L ) = T s e2 + L s e2‬‬ ‫)‪(5.10‬‬

‫) ‪= var( y T ) + L. var( e1‬‬

‫ﻫﺬﺍ ﺃﻳﻀﺎﹰ ﻳﺰﺩﺍﺩ ﻣﻊ ﺯﻳﺎﺩﺓ ﺃﻓﻖ ﺍﻟﺘﻨﺒﺆ‪.‬‬
‫ﺑﻄﺮﻳﻘﺔ ﺃﺧﺮﻯ‪ ،‬ﳝﻜﻦ ﺣﺴﺎﺏ ﺗﺒﺎﻳﻦ ﺍﻟﺘﻨﺒﺆ ﻭﻛﻤﺎ ﻳﻠﻲ ﺑﻌﺪ ﺣﺴﺎﺏ ﺃﺧﻄﺎﺀ‬
‫ﺍﻟﺘﻨﺒﺆ‪:‬‬
‫‪e1 = y T+ 1 - $y T +1‬‬
‫‪= ( y T + e T+ 1 ) - y T‬‬

‫‪e1 = e T +1‬‬
‫ﺑﺎﻟﺘﺎﱄ ﻳﻜﻮﻥ ﺗﺒﺎﻳﻨﻪ ﻣﻌﻄﻰ ﺑـ‪:‬‬
‫‪var( e1 ) = E( e 1 ) 2 = E( e T+1 ) 2 = se2‬‬ ‫)‪(5.11‬‬

‫‪139‬‬
 ‫‪e2 = y T+ 2 - $y T+ 2‬‬ ‫ﻭﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ‬
‫‪= ( y T+ 1 + e T+ 2 ) - y T‬‬
‫‪= y T + e T+ 1 + e T + 2 - y T‬‬

‫‪e 2 = e T+1 + e T+ 2‬‬

‫‪var( e2 ) = E( e 2 ) 2 = E( e T+1 + e T+ 2 ) 2‬‬ ‫ﺇﺫﺍ‬
‫ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻔﺮﺿﻴﺔ )‪ (5.2‬ﻭ)‪ (5.3‬ﺩﺍﺋﻤ‪‬ﺎ‬
‫‪var( e2 ) = 2 s e2‬‬ ‫)‪(5.12‬‬

‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫‪var( e L ) = L s‬‬ ‫‪2‬‬
‫‪e‬‬
‫)‪(5.13‬‬

‫ﻫﻮ ﻣﺘﺰﺍﻳﺪ ﻣﻊ ﺍﻟﺰﻣﻦ‪ .‬ﻧﻌﱪ ﻋﻦ ﻫﺬﺍ ﺍﳌﺸﻜﻞ ﺑﻌﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ‬

‫)‪ (Non-stationarity‬ﻣﻦ ﺣﻴﺚ ﺍﻟﺘﺒﺎﻳﻦ‪.‬‬

‫‪Stationarity‬‬ ‫‪ -2‬ﺍﻻﺳﺘﻘﺮﺍﺭﻳﺔ‪:‬‬
‫ﺗﻜﻮﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻣﺴﺘﻘﺮﺓ ﺃﻭ ﺳﺎﻛﻨﺔ ﺇﺫﺍ ﺗﺬﺑﺬﺑﺖ ﺣﻮﻝ ﻭﺳﻂ‬
‫ﺣﺴﺎﰊ ﺛﺎﺑﺖ‪ ،‬ﻣﻊ ﺗﺒﺎﻳﻦ ﻟﻴﺲ ﻟﻪ ﻋﻼﻗﺔ ﺑﺎﻟﺰﻣﻦ‪ .‬ﻭﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻨﻬﺎ ﺭﻳﺎﺿﻴﺎ ﲟﺎ‬
‫ﻳﻠﻲ ‪:‬‬
‫‪E( yt ) = m‬‬ ‫)‪(5.14‬‬

‫‪E( y t - m ) 2 = s2y = g 0 < ¥‬‬ ‫)‪(5.15‬‬

‫‪E( y t - k - m ) 2 = s 2y = g 0 < ¥‬‬

‫‪E ( y t - m )( y t - k - m ) = g‬‬ ‫‪k‬‬
‫)‪(5.16‬‬

‫ﻭ ‪ g k‬ﻟﻴﺲ ﻟﻪ ﻋﻼﻗﺔ ﺳﻮﻯ ﺑﺎﻟﻔﺮﻕ ﺑﲔ ‪ t‬ﻭ ‪ ، t - k‬ﺣﻴﺚ‬

‫‪g -k = g‬‬ ‫‪k‬‬

‫‪140‬‬
‫ﻛﻮﻥ ﻫﺬﻩ ﺍﳌﻘﺎﻳﻴﺲ ﺧﺎﺻﺔ ﺑﺎ‪‬ﺘﻤﻊ‪ ،‬ﳝﻜﻦ ﺣﺴﺎﺏ ﺗﻠﻚ ﺍﳋﺎﺻﺔ ﺑﺎﻟﻌﻴﻨﺔ‬
‫ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪ (1‬ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ‬
‫_‬
‫‪1 T‬‬ ‫)‪(5.17‬‬
‫= ‪m$ = y‬‬ ‫‪å yt‬‬
‫‪T t =1‬‬
‫‪ (2‬ﺍﻟﺘﺒﺎﻳﻦ‬
‫‪1 T‬‬ ‫_‬
‫)‪(5.18‬‬
‫= ‪g$ 0‬‬ ‫) ‪å( yt - y‬‬
‫‪2‬‬

‫‪T t =1‬‬
‫‪1 T‬‬ ‫_‬ ‫_‬
‫)‪(5.19‬‬
‫‪g$‬‬ ‫‪k‬‬ ‫=‬ ‫) ‪å ( y t - y )( y t- k - y‬‬
‫‪T t = k +1‬‬

‫ﺑﺎﻟﺮﺟﻮﻉ ﺇﱃ )‪ (5.1‬ﻭﻛﺘﺎﺑﺘﻪ ﰲ ﺍﻟﺸﻜﻞ ﺍﳉﺪﻳﺪ ﺍﻟﺘﺎﱄ ‪:‬‬

‫‪y t = fy t -1 + e t‬‬ ‫)‪(5.20‬‬

‫ﻳﻜﻮﻥ ﺍﻟﻨﻤﻮﺫﺝ )‪ (5.20‬ﻫﻮ ﻧﻔﺴﻪ )‪ (5.1‬ﳌﺎ ‪. f = 1‬‬

‫ﺭﻏﺒﺔ ﰲ ﺍﻟﺒﺤﺚ ﻋﻦ ﻣﺪﻯ ﺍﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﻨﻤﻮﺫﺝ ﻭﺑﺎﺳﺘﻌﻤﺎﻝ ﻓﻜﺮﺓ ﺍﳌﻌﺎﺩﻻﺕ‬
‫ﺍﻟﺘﻔﺎﺿﻠﻴﺔ ‪ ،Differential equations‬ﻭﺑﻌﺪ ﺇﳘﺎﻝ ﺍﻟﻌﻨﺼﺮ ﺍﻟﻌﺸﻮﺍﺋﻲ‪ ،‬ﻳﻜﻮﻥ ﺣﻞ‬
‫ﺍﻟﻌﻼﻗﺔ )‪ (5.20‬ﻣﺘﻤﺜﻞ ﰲ ﺍﳌﻌﺎﺩﻟﺔ ﺍﳌﺘﺠﺎﻧﺴﺔ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‪.‬‬

‫ﺍ ‪ -‬ﻣﻌﺎﺩﻟﺔ ﻣﺘﺠﺎﻧﺴﺔ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‪:‬‬

‫‪y t = fy t -1‬‬

‫‪t = 1 : y 1 = fy 0‬‬

‫‪t = 2 : y 2 = fy 1 = f ( fy 0 ) = f 2 y 0‬‬

‫‪t = 3: y 3 = fy 2 = f 3 y 0‬‬

‫‪t = j: y j = f j y 0‬‬ ‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬

‫‪141‬‬
 ‫ﻭﻣﻨﻪ‬
‫‪= f t y0‬‬ ‫)‪(5.21‬‬

‫‪f <1‬‬ ‫ﺣﱴ ﻳﻜﻮﻥ ﺍﳊﻞ ﻣﺴﺘﻘﺮﺍﹰ ﳚﺐ ﺃﻥ ﺗﻜﻮﻥ‪:‬‬

‫ﺭﺃﻳﻨﺎ ﳌﺎ ‪ f = 1‬ﺃﻥ ﺍﻟﻨﻤﻮﺫﺝ ﻳﻜﻮﻥ ﻏﲑ ﺛﺎﺑﺖ ﺍﻟﺘﺒﺎﻳﻦ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﻏﲑ‬
‫ﻣﺴﺘﻘﺮ‪‬ﺍ‪.‬‬
‫)‪(Back substitution‬‬ ‫ﻭﺑﺎﻟﺘﻌﻮﻳﺾ ﺍﳋﻠﻔﻲ‬ ‫)‪(5.20‬‬ ‫ﺑﺎﻟﺮﺟﻮﻉ ﺇﱃ ﺍﳌﻌﺎﺩﻟﺔ‬
‫ﳓﺼﻞ ﻋﻠﻰ ﻣﺎ ﻳﻠﻲ‪:‬‬
‫‪j- 1‬‬
‫‪y t = f j y t- j + å f i e t- i‬‬ ‫)‪(5.22‬‬
‫‪i=0‬‬

‫ﻣﻦ ﺃﻋﻼﻩ‬
‫‪t-1‬‬
‫‪y t = f t y0 + å f i e t- i‬‬ ‫)‪(5.23‬‬
‫‪i=0‬‬

‫‪ ، f‬ﻭﺗﺒﺎﻳﻦ‬ ‫‪=1‬‬ ‫ﲣﺘﻠﻒ )‪ (5.23‬ﻋﻦ )‪ (5.6‬ﰲ ﻛﻮﻥ ﺍﻷﺧﲑﺓ ﺗﻔﺘﺮﺽ ﺿﻤﻨﻴﺎ‬

‫)‪ (5.23‬ﻳﻜﻮﻥ‪:‬‬
‫‪var( y t ) = E [ y t - E ( y t ) ] 2‬‬
‫‪2‬‬
‫‪é t -1‬‬ ‫‪ù‬‬
‫‪= E êå f i e‬‬ ‫‪t -i ú‬‬
‫‪ë i =0‬‬ ‫‪û‬‬

‫ﺃﻳﻦ‬
‫‪E ( yt ) = f t y0‬‬ ‫)‪(5.24‬‬

‫‪var( y t ) = g 0 = E [ e t + fe t -1 + f 2 e t - 2 ... + f t -1 e 1 ] 2‬‬

‫) )‪= s 2 (1 + f 2 + f 4 + ... f 2 ( t - 1‬‬

‫‪142‬‬
‫)ﺣﱴ ﻳﻜﻮﻥ ﺍﻟﺘﺒﺎﻳﻦ ﻣﻌﻘﻮﻻﹰ ﺃﻱ ﻏﲑ‬ ‫‪f <1‬‬ ‫ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﺸﺮﻁ ﺍﻟﺴﺎﺑﻖ‬
‫ﺳﺎﻟﺐﹴ( ﳓﺼﻞ ﻋﻠﻰ‪:‬‬
‫‪s2‬‬ ‫)‪(5.25‬‬
‫= ‪g0‬‬
‫‪1 - f2‬‬
‫ﻛﻮﻥ ﻣﺎﺑﲔ ﺍﻟﻘﻮﺳﲔ ﳝﺜﻞ ﳎﻤﻮﻉ ﻋﻨﺎﺻﺮ ﻣﺘﻮﺍﻟﻴﺔ ﻫﻨﺪﺳﻴﺔ‪ ،‬ﺃﺳﺎﺳﻬﺎ‬
‫= ‪ r‬ﻭﺣﺪﻫﺎ ﺍﻷﻭﻝ ‪ a = 1‬ﺃﻱ‬ ‫‪f2‬‬
‫) ‪a ( 1 - rj‬‬ ‫‪1‬‬
‫=‪S‬‬ ‫=‬
‫‪1- r‬‬ ‫‪1 - f2‬‬

‫ﺣﻴﺚ ‪ r j ® 0‬ﳌﺎ ‪ j ® ¥‬ﺇﺫﺍ ﻛﺎﻥ ‪ . r < 1‬ﻣﻦ ﻫﻨﺎ ﻧﺴﺘﻨﺘﺞ ﺃﻧﻪ ﳚﺐ‬
‫ﺃﻥ ﺗﻜﻮﻥ ‪ f < 1‬ﺣﱴ ﻳﻜﻮﻥ ﺍﻟﻨﻤﻮﺫﺝ )‪ (5.20‬ﻣﺴﺘﻘﺮ‪‬ﺍ‪.،‬‬

‫ﺏ‪ -‬ﻣﻌﺎﺩﻟﺔ ﻏﲑ ﻣﺘﺠﺎﻧﺴﺔ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‪:‬‬

‫ﻫﺬﻩ ﺍﻷﺧﲑﺓ ﺗﻜﺘﺐ ﻭﺑﻌﺪ ﺇﳘﺎﻝ ﺍﳋﻄﺄ ﺍﻟﻌﺸﻮﺍﺋﻲ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪y t = f y t-1 + d‬‬ ‫)‪(5.26‬‬

‫ﺣﻞ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻟﺔ ﻳﺘﻤﺜﻞ ﰲ‬

‫‪t -1‬‬
‫‪y t = f t y0 + ( å f i )d‬‬
‫‪i=0‬‬

‫ﻳﻜﻮﻥ ﻫﺬﺍ ﺍﳊﻞ ﻣﻌﻘﻮﻻﹰ ﳌﺎ ﺗﻜﻮﻥ‪:‬‬ ‫‪f < 1:‬‬ ‫ﻭﺗﺒﺎﻳﻨﻬﺎ ﺛﺎﺑﺘ‪‬ﺎ ﻭﻣﺴﺎﻭﻳ‪‬ﺎ‬
‫‪s2‬‬
‫= ‪g0‬‬
‫‪1 - f2‬‬
‫ﻳﻜﻮﻥ ﻭﻓﻖ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺘﻨﺒﺆ ﻣﻌﻄﻰ ﲟﺎ ﻳﻠﻲ‪:‬‬
‫^‬
‫) ‪y T +1 = E ( y T +1 / y T , . . . , y 1‬‬
‫) ‪= E ( fy T + d + e T +1‬‬
‫‪= fyT + d‬‬

‫‪143‬‬
 ‫) ‪$y T+ 2 = E ( y T+ 2 / y T ,..., y 1‬‬
‫‪= E ( fy T + 2 + d + e T + 2‬‬
‫‪= E ( f ( fy T + d + e T + 1 ) + d + e T + 2‬‬
‫‪= f 2 y T + fd + d‬‬
‫ﺣﻴﺚ ﺗﻜﻮﻥ ﺗﻮﻗﻌﺎﺕ ﺍﳌﻘﺎﺩﻳﺮ ﺍﻷﺧﺮﻯ ﻣﻌﺪﻭﻣﺔ‪.‬‬
‫‪$y T+ 3 = f 3 y T + f 2 d + fd + d‬‬ ‫ﺑﺎﻟﺘﺸﺎﺑﻪ‬
‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫‪L -1‬‬
‫‪$y T + L = f L y T + ( å f i ) d‬‬
‫‪i=0‬‬

‫ﺣﻴﺚ ‪ L‬ﳝﺜﻞ ﺃﻓﻖ ﺍﻟﺘﻨﺒﺆ‪ ،‬ﻭﰲ ﺣﺎﻟﺔ ‪ f = 1‬ﻓﺈﻥ‬

‫‪$y T + L = y‬‬ ‫‪T‬‬ ‫‪+ Ld‬‬

‫ﺍﻟﻮﺍﺭﺩ ﰲ ﺍﻟﺒﺎﺏ‬ ‫)‪(3.9‬‬ ‫ﻫﻮ ﺗﻘﺮﻳﺒ‪‬ﺎ‪ ،‬ﻧﻔﺲ ﺗﻨﺒﺆ ﳕﻮﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺍﶈﻠﻲ‬
‫ﺍﻷﻭﻝ ﺇﺫﺍ ﻛﺎﻥ ‪. d = y t - y t -1‬‬
‫‪E ( e L ) = Ls e2‬‬ ‫ﻳﻜﻮﻥ ﺗﺒﺎﻳﻦ ﺧﻄﺄ ﺍﻟﺘﻨﺒﺆ‬
‫ﻛﻤﺎ ﻳﻜﻮﻥ ﺍﻟﻨﻤﻮﺫﺝ ﻭﻓﻖ ﺍﳊﺎﻟﺔ ﺍﻷﺧﲑﺓ ﻏﲑ ﻣﺴﺘﻘﺮﹴ )ﳕﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ‬
‫ﺍﻟﻌﺸﻮﺍﺋﻲ(‪ ،‬ﻭﺃﺳﺒﺎﺏ ﻗﺪ ﺗﺘﻤﺜﻞ ﰲ ﻭﺟﻮﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﻭﺍﻻﲡﺎﻩ‬
‫ﺍﻟﻌﺎﻡ ﰲ ﺍﻟﺘﺒﺎﻳﻦ )ﺗﺒﺎﻳﻦ ﻏﲑ ﺛﺎﺑﺖ(‪ .‬ﻟﻠﺘﺨﻠﺺ ﻣﻦ ﻣﺸﻜﻞ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ‪ ،‬ﳚﺐ‬
‫ﺃﻭﻻﹰ ﻣﻌﺮﻓﺔ ﻣﺴﺒﺒﺎﺗﻪ‪ ،‬ﰒ ﳏﺎﻭﻟﺔ ﺇﺯﺍﻟﺘﻬﺎ ﺑﺈﺣﺪﻯ ﺍﻟﻄﺮﻕ ﺍﳌﺬﻛﻮﺭﺓ ﰲ ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ‪،‬‬
‫ﺑﻌﺪ ﺗﻄﺒﻴﻖ ﺍﻟﺘﻘﻨﻴﺔ ﳌﺮﺓ ﺃﻭ ﻣﺮﺗﲔ‪ .‬ﻣﻦ ﻫﺬﻩ ﺍﻟﻄﺮﻕ‪ ،‬ﺗﻘﻨﻴﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ‬
‫ﺍﻷﻭﱃ )ﻹﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ(‪ ،‬ﻭ ﻣﻦ ﺍﻟﺪﺭﺟﺔ )‪ (p‬ﻻﺳﺘﺒﻌﺎﺩ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ‬
‫ﻣﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﺑﻴﻨﻤﺎ ﳝﻜﻦ ﺇﺯﺍﻟﺔ ﺍﻟﺴﺒﺐ ﺍﻟﺜﺎﻟﺚ ﺑﻄﺮﻳﻘﺔ ﺑﻮﻛﺲ‪-‬ﻛﻮﻛﺲ‬
‫ﰲ ﺷﻜﻠﻬﺎ ﺍﻟﺒﺴﻴﻂ‪ ،‬ﻭﺍﳌﺘﻤﺜﻞ ﰲ ﺇﺩﺧﺎﻝ ﺍﻟﻠﻮﻏﺎﺭﻳﺘﻢ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ‪.‬‬

‫‪144‬‬
‫ﻧﺮﻯ ﺿﻤﻦ ﻫﺬﺍ ﺍﻟﺒﺎﺏ‪ ،‬ﻛﻴﻔﻴﺔ ﻛﺸﻒ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﻭﺫﻟﻚ ﺑﺎﺳﺘﻌﻤﺎﻝ‬
‫ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻣﻊ ﺍﺧﺘﺒﺎﺭ ﺑﺎﺭﺗﻠﺖ )‪ (Bartlett‬ﻭﺍﳌﺴﻤﻰ ﺑـ ‪(Rule of‬‬
‫)‪ ،thumb‬ﺍﻟﺬﻳ‪‬ﻦ ﺳﻨﺘﻌﺮﻑ ﻋﻠﻴﻬﻤﺎ ﻓﻴﻤﺎ ﺑﻌﺪ‪ ،‬ﺣﻴﺚ ﻣﻌﺎﻣﻼﺕ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺗﺘﺠﻪ‬
‫ﺑﻌﻴﺪﺍﹰ ﻋﻦ ﺍﻻﻧﻌﺪﺍﻡ ﰲ ﺁﺟﺎﻝ ﳏﺪﺩﺓ‪ ،‬ﺑﺪﻭﻥ ﺍﻻﻛﺘﺮﺍﺙ ﺑﺎﻟﺸﺮﻭﻁ ﺍﳋﺎﺻﺔ ﲟﻌﻠﻤﺎﺕ‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﻭﺍﻟﱵ ﺭﺃﻳﻨﺎﻫﺎ ﺳﺎﺑﻘﺎ‪ ،‬ﻭﺍﻟﱵ ﺳﻨﻄﺮﺣﻬﺎ ﻓﻴﻤﺎ ﺑﻌﺪ ﻋﻨﺪ ﺗﻨﺎﻭﻟﻨﺎ ﻣﻮﺿﻮﻉ‬
‫ﲢﺪﻳﺪ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳋﻄﻴﺔ ﻟﻠﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ‪.‬‬
‫ﻛﻤﺎ ﻧﺴﺘﻌﺮﺽ ﻻﺣﻘﺎﹰ ﺍﺧﺘﺒﺎﺭﺍﺕ ﺩﻳﻜﻲ ﻓﻮﻟﺮ ﻟﺘﺤﺪﻳﺪ ﻣﺪﻯ ﺍﺳﺘﻘﺮﺍﺭﻳﺔ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻭﻣﻦ ﰒ ﺍﻟﺴﺒﻴﻞ ﺍﻷﻧﺴﺐ ﻟﺘﺤﻮﻳﻠﻬﺎ ﺇﱃ ﺳﻠﺴﻠﺔ ﺳﺎﻛﻨﺔ‪.‬‬

‫‪ - 3‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‪:‬‬

‫ﺗﻮﺿﺢ ﻫـﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﳌﻮﺟﻮﺩ ﺑﲔ ﺍﳌﺸﺎﻫـﺪﺍﺕ ﰲ ﻓﺘﺮﺍﺕ‬
‫ﳐﺘﻠﻔﺔ‪ ،‬ﻭﻫﻲ ﺫﺍﺕ ﺃﳘﻴﺔ ﺑﺎﻟﻐﺔ ﰲ ﺇﺑﺮﺍﺯ ﺑﻌﺾ ﺍﳋﺼﺎﺋﺺ ﺍﳍﺎﻣﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬
‫ﻳﻌﺮﻑ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ) ‪ ( k‬ﻛﺎﻟﺘﺎﱄ‪:‬‬
‫) ‪cov( y t , y t - k‬‬
‫= ‪rk‬‬
‫) ‪var( y t ) var( y t - k‬‬

‫) ‪E ( y t - m y )( y t - k - m y‬‬
‫= ‪rk‬‬
‫‪E ( y t - m y ) 2 . E ( y t- k - m y ) 2‬‬

‫‪g‬‬ ‫‪gk‬‬
‫=‬ ‫‪k‬‬
‫=‬
‫‪s .s‬‬ ‫‪2‬‬
‫‪y‬‬
‫‪2‬‬
‫‪y‬‬
‫‪s 2y‬‬

‫‪gk‬‬
‫=‬ ‫ﺇﺫﺍ‬
‫‪g0‬‬

‫‪145‬‬
‫ﺗﻜﻮﻥ ﻣﻘﺪ‪‬ﺭﺍﺕ ﻫﺬﻩ ﺍﻟﺘﺒﺎﻳﻨﺎﺕ ﻭﺍﻟﺘﺒﺎﻳﻨﺎﺕ ﺍﳌﺸﺘﺮﻛﺔ ﰒ ﻣﻌﺎﻣﻼﺕ ﺍﻻﺭﺗﺒﺎﻁ‬
‫ﺍﻟﺬﺍﰐ ﺍﳋﺎﺹ ﺑﺎﻟﻌﻴﻨﺔ‬
‫‪g$ k‬‬
‫= ‪r$ k‬‬ ‫‪= rk‬‬
‫‪g$ 0‬‬
‫‪1 T‬‬
‫ﺣﻴﺚ‬
‫_‬ ‫_‬
‫‪g$‬‬ ‫‪k‬‬ ‫=‬ ‫‪å ( y t - y )( y t- k - y ) = c k‬‬
‫‪T t = k +1‬‬
‫‪1 T‬‬ ‫_‬
‫= ‪g$ 0‬‬ ‫‪å ( y t - y )2 = c0‬‬
‫‪T t=1‬‬
‫ﺃﻱ‬
‫‪T‬‬ ‫_‬ ‫_‬

‫‪å ( y - y)( y‬‬

‫‪t‬‬ ‫‪t -k‬‬ ‫)‪- y‬‬
‫‪r̂k = t = k +1‬‬ ‫‪T‬‬ ‫_‬ ‫‪= rk‬‬ ‫)‪(5.27‬‬
‫)‪å ( y - y‬‬
‫‪t =1‬‬
‫‪t‬‬
‫‪2‬‬

‫‪1 T‬‬
‫ﺃﻳﻦ‬
‫_‬
‫= ‪y‬‬ ‫‪åyt‬‬
‫‪T t =1‬‬
‫ﻣﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫‪r$ k‬‬ ‫ﻭﲢﻘﻴﻖ ﺃﺧﺮ ﻝ‬
‫_‬
‫_‬ ‫‪c‬‬
‫‪rk = k‬‬
‫‪c0‬‬
‫ﺣﻴﺚ‬
‫_‬
‫‪1‬‬ ‫‪T‬‬ ‫_‬ ‫_‬
‫= ‪Ck‬‬
‫‪T -k‬‬
‫‪å (y‬‬
‫‪t = k +1‬‬
‫‪t‬‬ ‫) ‪- y )( y t - k - y‬‬

‫‪r k = r-k‬‬ ‫ﳘﺎ ﻣﺘﺴﺎﻭﻳﲔ ﺗﻘﺮﻳﺒﺎ ﰲ ﺍﻟﻌﻴﻨﺎﺕ ﺍﻟﻜﺒﲑﺓ‪ ،‬ﻛﻤﺎ ﺑﺄﻥ‬

‫‪ - 2‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﳋﻄﻴﺔ ﻟﻠﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ‪:‬‬

‫ﺑﻌﺪ ﺍﳍﺪﻑ ﺍﻟﺘﻌﻠﻴﻤﻲ‪ ،‬ﻳﻜﻮﻥ ﺍﳍﺪﻑ ﺍﻷﺳﺎﺳﻲ ﳍﺬﻩ ﺍﻟﺪﺭﺍﺳﺔ ﻣﺘﻤﺜﻼﹰ ﰲ‬
‫ﺑﻨﺎﺀ ﳕﺎﺫﺝ ﺧﻄﻴﺔ ﻟﻠﻈﺎﻫﺮﺓ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻭﺍﺳﺘﻌﻤﺎﳍﺎ ﰲ ﻣﻴﺪﺍﻥ ﺍﻟﺘﻨﺒﺆ ﻋﻠﻰ ﺃﺳﺎﺱ ﺗﻔﺴﲑ‬

‫‪146‬‬
‫ﺳﻠﻮﻙ ﻣﺘﻐﲑ ﻣﺎ ﻣﻦ ﺧﻼﻝ ﺧﺼﺎﺋﺼﻪ ﺍﻟﺒﺎﺭﺯﺓ ﻭﺍﳌﺘﻤﺜﻠﺔ ﰲ ﻣﺎﺿﻲ ﻫﺬﺍ ﺍﳌﺘﻐﲑ‬
‫ﺍﳌﺪﺭﻭﺱ‪.‬‬
‫ﻳﻘﺴﻢ ﻫﺬﺍ ﺍﳌﻮﺿﻮﻉ ﺣﺴﺐ ﺑﻮﻛﺲ‪-‬ﺟﺎﻧﻜﻴﻨﺲ ﺇﱃ ﺛﻼﺛﺔ ﻣﺮﺍﺣﻞ ﺭﺋﻴﺴﻴﺔ‬
‫ﻫﻲ‪:‬‬
‫»‪«Identification‬‬ ‫ﻣﺮﺣﻠﺔ ﲢﺪﻳﺪ ﺍﻟﻨﻤﻮﺫﺝ‬ ‫‪(1‬‬

‫»‪«Estimation‬‬ ‫ﻣﺮﺣﻠﺔ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ‬ ‫‪(2‬‬

‫»‪«Diagnostic checking‬‬ ‫ﻣﺮﺣﻠﺔ ﺍﻻﺧﺘﺒﺎﺭ‬ ‫‪(3‬‬

‫‪ - 1.2‬ﻣﺮﺣﻠﺔ ﲢﺪﻳﺪ ﺍﻟﻨﻤﻮﺫﺝ‬

‫ﻳﺘﻢ ﰲ ﻫﺬﻩ ﺍﳌﺮﺣﻠﺔ ﺍﻟﺘﻌﺮﻑ ﻋﻠﻰ ﳕﻮﺫﺝ ﻳﻔﺴﺮ ﺳﻠﻮﻙ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻣﻦ‬
‫ﺧﻼﻝ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻭﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ‪ ،‬ﺍﻟﱵ ﺳﻨﺴﺘﻌﺮﺿﻬﺎ‬
‫ﻻﺣﻘﺎﹰ‪ .‬ﺗﺴﺘﻌﻤﻞ ﻫﺬﻩ ﺍﻟﺪﻭﺍﻝ ﻻﺳﺘﻨﺒﺎﻁ ﺍﳋﺼﺎﺋﺺ ﺍﳍﺎﻣﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﺍﻟﱵ‬
‫ﺗﺴﻤﺢ ﺑﺘﺤﺪﻳﺪ ﺍﻟﻨﻤﻮﺫﺝ ﺃﻭ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﻼﺋﻤﺔ ﺍﳌﻨﺘﻤﻴﺔ ﺇﱃ ﳎﻤﻮﻋﺔ ﳕﺎﺫﺝ ﺑﻮﻛﺲ‪-‬‬
‫ﺟﺎﻧﻜﻴﻨﺰ ﺍﳌﺘﻤﺜﻠﺔ ﰲ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‪ ،‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ‪ ،‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ‬
‫ﻭ ﺍﳌﺨﺘﻠﻄﺔ ﺍﳌﺮﻛﺒﺔ‪.‬‬

‫)‪MA(q‬‬
‫‪ - 1.1.2‬ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‬
‫‪1‬‬

‫ﺃﻭﻝ ﺍﻟﺘﺴﺎﺅﻻﺕ ﺍﻟﱵ ﳝﻜﻦ ﻃﺮﺣﻬﺎ ﺍﻵﻥ ﺣﻮﻝ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﻫﻲ ﻣﺎ‬
‫ﺷﻜﻠﻬﺎ‪ ،‬ﻭﻣﺎ ﺷﻜﻞ ﺩﺍﻟﺔ ﺍﺭﺗﺒﺎﻃﻬﺎ ﺍﻟﺬﺍﺗﻴﺔ؟‪.‬‬
‫ﻳﻜﺘﺐ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﰲ ﺷﻜﻞ ﺧﻄﻲ ﻋﺎﻡ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y t = m + e t + q 1 e t - 1 + q 2 e t - 2 +... + q q e t - q .‬‬ ‫)‪(5.28‬‬

‫) ‪1 - Moving Average of Order(q‬‬

‫‪147‬‬
 ‫ﻫﻮ ﻣﻦ ﺍﻟﺪﺭﺟﺔ )‪ ،(q‬ﺍﳌﻘﺎﺳﺔ ﺑﻌﺪﺩ ﻣﻌﻠﻤﺎﺗﻪ‪.‬‬
‫ﻻﺳﺘﻨﺒﺎﻁ ﺍﳋﺼﺎﺋﺺ ﺍﳉﻮﻫﺮﻳﺔ ﻟﻠﻨﻤﻮﺫﺝ ﻧﻈﺮﻳ‪‬ﺎ ﻧﻌﻴﺪ ﻃﺮﺡ ﺍﻟﻔﺮﺿﻴﺎﺕ‬
‫ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪E( e t ) = 0‬‬

‫‪E ( e t ) 2 = s e2‬‬

‫‪E ( e t e t - k ) = 0 "k ¹ 0‬‬

‫ﻣﻦ ﺷﺮﻭﻁ ﺍﺳﺘﻘﺮﺍﺭ ﻫﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ ﺃﻥ ﻳﻜﻮﻥ ﻭﺳﻄﻬﺎ ﻣﺴﺘﻘﻼﹰ ﻋﻦ ﺍﻟﺰﻣﻦ‬

‫ﻭﺗﺒﺎﻳﻨﻬﺎ ‪‬ﺎﺋﻲ‪ .‬ﻫﺬﺍ ﻳﻌﲏ ﺃﻥ ﻭﺳﻄﻬﺎ ﲢﺖ ﻫﺬﻩ ﺍﻟﻔﺮﺿﻴﺎﺕ ﻳﺴﺎﻭﻱ ‪ m‬ﻭﻫﻮ‬
‫ﻣﺴﺘﻘﻼﹰ ﻋﻦ ﺍﻟﺰﻣﻦ ﺃﻱ‬
‫‪E( y t ) = m‬‬
‫ﺑﻴﻨﻤﺎ ﺗﺒﺎﻳﻨﻬﺎ ﻫﻮ‬
‫‪var( y t ) = g 0 = E ( y t - m ) 2‬‬
‫‪= E ( e t + q 1 e t - 1 + q 2 e t - 2 +... + q q e t - q ) 2‬‬

‫) ‪= s e2 ( 1 + q 1 + q 2 +... + q q‬‬

‫ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻔﺮﺿﻴﺔ‬

‫‪E ( e t e t - k ) = 0"k ¹ 0‬‬

‫ﻭﻫﻮ ﻣﻘﺪﺍﺭ ‪‬ﺎﺋﻲ‪ ،‬ﺃﻱ‬

‫‪q‬‬
‫‪åqi < ¥‬‬
‫‪2‬‬

‫‪i=1‬‬

‫ﻓﺈﺫﺍ ﻛﺎﻧﺖ ‪ q = 1‬ﺃﻱ ﳕﻮﺫﺝ )‪ ،MA(1‬ﻳﻜﺘﺐ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪y t = m + e t + q 1e t-1‬‬

‫‪148‬‬
 E( y t ) = m

var( y t ) = g 0 = E ( y t - m ) 2

= E ( e t + q 1 e t -1 ) 2

= E ( e 2t + 2 qe t e t - 1 + q 2 e 2t - 1 )

= s e2 ( 1 + q 2 )

g 1 = E ( y t - m )( y t - 1 - m )

g 1 = ( e t + qe t -1 )( e t -1 + qe t - 2 )

= qs e2
‫ ﻷﻥ‬g 1 = g -1 ‫ﻧﺸﲑ ﻫﻨﺎ ﺃﻥ‬
g -1 = E ( y t - 1 - m )( y t - m )

= qs e2

g 2 = E ( y t - m )( y t - 2 - m )

g 2 = ( e t + qe t - 1 )( e t - 2 + qe t - 3 )

=0
g k = 0 "k > 1 ‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
:‫ﳝﻜﻦ ﺗﻠﺨﻴﺺ ﻫﺬﻩ ﺍﻟﻨﺘﻴﺠﺔ ﻓﻴﻤﺎ ﻳﻠﻲ‬
‫ ﻏﲑ ﻣﻌﺪﻭﻣﺎ‬g k ‫ﻳﻜﻮﻥ‬ k £q ‫ﳌﺎ ﻳﻜﻮﻥ‬
‫ﻣﻌﺪﻭﻣﺎﹰ‬ gk ‫ﻳﻜﻮﻥ‬ k >q ‫ﳌﺎ ﻳﻜﻮﻥ‬

149
 ‫ﺗﺒﻌﺎﹼ ﳌﺎ ﺳﺒﻖ‪ ،‬ﺗﻜﻮﻥ ﺩﺍﻟﺔ ﺍﺭﺗﺒﺎﻁ ﺫﺍﰐ ﻟﻨﻤﻮﺫﺝ ﻣﺎ ﻫﻲ‬
‫‪gk‬‬
‫= ‪rk‬‬
‫‪g0‬‬

‫‪k =1‬‬ ‫ﺇﺫﺍ ﻓﻠﻤﺎ‬

‫‪g1‬‬
‫= ‪r1‬‬
‫‪g0‬‬

‫‪qs e2‬‬ ‫‪q‬‬

‫‪= 2‬‬ ‫=‬
‫) ‪se ( 1 + q ) ( 1 + q2‬‬
‫‪2‬‬

‫‪k =2‬‬ ‫ﻭ ﳌﺎ‬

‫‪g2‬‬
‫‪r2 = 0‬‬ ‫= ‪r2‬‬
‫‪g0‬‬

‫ﺇﺫﺍ ﻛﺎﻧﺖ‬ ‫) ‪MA ( 1‬‬ ‫ﺍﺣﺴﺐ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻨﻈﺮﻳﺔ ﻟﻨﻤﻮﺫﺝ‬ ‫‪:24‬‬ ‫ﻣﺜﺎﻝ‬
‫‪. q = 0. 7‬‬
‫ﳓﺼﻞ ﻋﻠﻰ‪:‬‬ ‫‪g‬‬ ‫‪k‬‬ ‫ﻭ‬ ‫‪g 2 ،g 1 ،g‬‬ ‫‪0‬‬
‫ﺇﺫﺍ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻌﻼﻗﺎﺕ ﺍﻟﺴﺎﺑﻘﺔ ﻝ‬

‫‪y t = m + e t + 0. 7e t - 1‬‬

‫) ‪g 0 = s e2 ( 1 + q 2‬‬ ‫ﻣﻨﻪ‬
‫‪g 0 = 1. 49s e2‬‬

‫‪g 1 = qs e2 = 0. 7s e2‬‬

‫‪g2 = 0‬‬

‫‪150‬‬
 ‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫‪gk =0‬‬
‫ﻳﺘﻢ ﺣﺴﺎﺏ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪g0‬‬
‫= ‪r0‬‬ ‫‪=1‬‬
‫‪g0‬‬
‫‪g1‬‬ ‫‪0. 7s e2‬‬
‫= ‪r1‬‬ ‫=‬ ‫‪= 0. 47‬‬
‫‪g 0 1. 49 s e2‬‬

‫‪rk = 0‬‬ ‫ﻓﺈﻥ‬ ‫‪k >1‬‬ ‫ﳌﺎ ﺗﻜﻮﻥ‬

‫ﺗﻜﻮﻥ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻨﻤﻮﺫﺝ ﲟﻌﻠﻤﺔ ﻣﻮﺟﺒﺔ ﰒﹼ ﺳﺎﻟﺒﺔ ﳑﺜﻠﺔ ﰲ ﻫﺬﻳﻦ‬
‫ﺍﻟﺸﻜﻠﲔ‪:‬‬
‫ﺍﻟﺸﻜﻞ )‪ :(22‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟـ)‪ MA(1‬ﺑـ‪.q = 0. 7‬‬

‫ﻭﻣﻨﻪ‬ ‫‪g 1 = -0. 7s e2‬‬ ‫ﺑﻴﻨﻤﺎ ﺇﺫﺍ ﻛﺎﻧﺖ ‪ q = -0. 7‬ﻓﺈﻥ‬

‫‪g 1 -0. 7s e2‬‬
‫= ‪r1‬‬ ‫=‬ ‫‪= -0. 47‬‬
‫‪g 0 1. 49s e2‬‬

‫‪151‬‬
 ‫ﺍﻟﺸﻜﻞ )‪ :(23‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟـ)‪ MA(1‬ﺑـ‪. q = -0. 7‬‬

‫ﻳﻜﻮﻥ ﳑﺜﻼ‬ ‫)‪MA(2‬‬ ‫ﺃﻣﺎ ﳕﻮﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻟﺜﺎﻧﻴﺔ‬

‫ﺑﺎﳌﻌﺎﺩﻟﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪y t = m + e t + q 1 e t -1 + q 2 e t - 2‬‬

‫ﺑﻴﻨﻤﺎ ﺗﺒﺎﻳﻨﺎﺗﻪ ﺗﻜﻮﻥ ﻣﻌﻄﺎﺓ ﺑـ‪:‬‬ ‫‪E( y t ) = m‬‬ ‫ﺣﻴﺚ‬

‫) ‪g 0 = s e2 ( 1 + q 12 + q 22‬‬

‫) ‪g 1 = q 1 s e2 ( 1 + q 2‬‬

‫‪g 2 = q 2 s e2‬‬

‫‪g3 = 0‬‬
‫ﻣﻨﻬﺎ ﻧﺴﺘﻄﻴﻊ ﺣﺴﺎﺏ ﻭﺭﺳﻢ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬

‫‪152‬‬
 ‫‪g1‬‬ ‫) ‪q 1 s e2 ( 1 + q 2‬‬
‫= ‪r1‬‬ ‫=‬
‫) ‪g 0 s e2 ( 1 + q 12 + q 22‬‬

‫) ‪q1 ( 1 + q 2‬‬
‫=‬
‫) ‪( 1 + q 12 + q 22‬‬

‫‪q2‬‬
‫= ‪r2‬‬
‫) ‪( 1 + q 12 + q 22‬‬

‫ﺗﺼﺒﺢ ﺍﻟﺪﺍﻟﺔ ﻣﻌﻄﺎﺓ ﺑﺎﻟﻌﻼﻗﺎﺕ‬ ‫‪q 2 = 0. 5‬‬ ‫ﻭ‬ ‫‪q 1 = 0. 4‬‬ ‫ﻓﺈﺫﺍ ﻛﺎﻧﺖ‬
‫ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫) ‪g 0 = s e2 ( 1 + q12 + q 22‬‬

‫‪= 1. 41s e2‬‬

‫‪g 1 = 0. 4 s e2 ( 1 + 0. 5 ) = 0. 6s e2‬‬

‫‪g 2 = 0. 5s e2‬‬

‫‪g3 = 0‬‬

‫‪k >2‬‬ ‫ﺗﻜﻮﻥ ﳌﺎ‬

‫‪g‬‬ ‫‪k‬‬ ‫‪= 0‬‬

‫‪r 1 = 0. 43‬‬ ‫ﻣﻨﻪ‬

‫‪r 2 = 0. 35‬‬

‫‪k >2‬‬ ‫ﳌﺎ‬ ‫‪rk = 0‬‬ ‫ﻣﻨﻪ‬

‫‪153‬‬
 ‫)‪.MA(2‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(24‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻨﻤﻮﺫﺝ‬

‫ﻓﺈﻥ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺗﺄﺧﺬ ﺍﻟﺸﻜﻞ‪:‬‬ ‫‪q 2 = -0. 5‬‬ ‫ﺑﻴﻨﻤﺎ ﺇﺫﺍ ﻛﺎﻧﺖ ﻣﺜﻼ‬

‫)‪.MA(2‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(25‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻨﻤﻮﺫﺝ‬

‫ﻋﻠﻰ ﺃﺳﺎﺱ ﻣﺎ ﺳﺒﻖ‪ ،‬ﳝﻜﻦ ﺍﻟﻘﻮﻝ ﺃﻥ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺗﺒﺘﺮ )ﺗﻨﻌﺪﻡ(‬
‫ﻣﺒﺎﺷﺮﺓ ﺑﻌﺪ ﺍﻟﺪﺭﺟﺔ )‪ .(q‬ﻓﺈﺫﺍ ﻛﺎﻥ ﳕﻮﺫﺝ ﻋﺒﺎﺭﺓ ﻋﻦ )‪ MA(1‬ﻓﺈﻥ ‪، r 2 = 0‬‬

‫‪154‬‬
‫ﺑﻴﻨﻤﺎ ﺇﺫﺍ ﻛﺎﻥ )‪ MA(2‬ﻓﺈﻥ ‪ r 3 = 0‬ﻭﻫﻜﺬﺍ‪ .‬ﻣﻨﻪ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬
‫ﺍﳌﻮﺍﻟﻴﺔ ﲤﺜﻞ ﳕﻮﺫﺝ )‪ MA(3‬ﺑـ ‪ q 1 = 0‬ﻭ ‪ q 2 = 0‬ﺑﻴﻨﻤﺎ ‪ q‬ﻣﻮﺟﺒﺔ‪ ،‬ﻭﻫﻜﺬﺍ‪.‬‬
‫‪3‬‬

‫)‪MA(3‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(26‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳌﻮﺍﻟﻴﺔ ﲤﺜﻞ ﳕﻮﺫﺝ‬

‫‪AR(p):‬‬ ‫‪ -2.1.2‬ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ‬

‫‪2‬‬

‫ﻣﺎﺯﻟﻨﺎ ﺩﺍﺋﻤﺎ ﰲ ﻣﺮﺣﻠﺔ ﺍﻟﺘﻌﺮﻑ ﻋﻠﻰ ﳕﻮﺫﺝ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻣﻦ ﺧﻼﻝ‬

‫ﺍﻟﻨﻈﺮ ﺇﱃ ﻭﺳﻄﻬﺎ ﻭﺗﺒﺎﻳﻨﻬﺎ‪ ،‬ﻓﻴﻔﺴﺮ ﰲ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺘﻐﲑ ﺍﻟﺘﺎﺑﻊ‪ ،‬ﺍﳌﻤﺜﻞ‬
‫ﻟﻠﻈﺎﻫﺮﺓ ﺍﳌﺪﺭﻭﺳﺔ ﺑﻮﺍﺳﻄﺔ ﻣﺎﺿﻴﻪ ﻓﻘﻂ‪ ،‬ﻭﺍﻟﺬﻱ ﳝﺜﻞ ﺳﻠﻮﻛﻪ ﰲ ﺍﳌﺎﺿﻲ‪ ،‬ﻭﻳﺸﺎﺭ‬
‫ﺇﻟﻴﻪ ﺑﺎﻟﺮﻣﺰ )‪ ،AR(p‬ﻭﻳﻜﺘﺐ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y t = d + f 1 y t - 1 + f 2 y t - 2 + ... + f p y t - p + e t‬‬
‫)‪(5.29‬‬

‫ﲤﺜﻞ ﺩﺭﺟﺔ ﺍﻟﻨﻤﻮﺫﺝ‪.‬‬ ‫‪p‬‬ ‫ﺣﻴﺚ‬

‫ﻛﻤﺎ ﳝﻜﻦ ﻛﺘﺎﺑﺔ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺑﻌﺪ ﺇﺩﺧﺎﻝ ﻓﻜﺮﺓ ﻣﻌﺎﻣﻞ ﺍﻟﺘﺄﺧﲑ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫)‪2-Autoregressive of Order (p‬‬

‫‪155‬‬
 ‫‪y t = d + f 1 L 1 y t + f 2 L 2 y t +... + f p L p y t + e t‬‬
‫‪( 1 - f 1 L 1 - f 2 L 2 -... - f p L p ) y t = d + e t‬‬

‫‪F( L ) y t = d + e t‬‬

‫‪y t = F -1 ( L ) d + F -1 ( L ) e t‬‬
‫( ‪.F‬‬ ‫ﺍﳊﺪﻭﺩ ) ‪L‬‬ ‫ﺃﻳﻦ ) ‪ F -1 ( L‬ﻫﻮ ﻣﻘﻠﻮﺏ ﺃﻭ ﻣﻌﻜﻮﺱ ﻛﺜﲑ‬

‫ﺧﺼﺎﺋﺺ ﺍﻟﻨﻤﻮﺫﺝ‪:‬‬
‫ﻋﻠﻰ ﻏﺮﺍﺭ ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‪ ،‬ﳓﺎﻭﻝ ﺍﻟﻨﻈﺮ ﰲ ﺧﺼﺎﺋﺺ ﺍﻟﻨﻤﺎﺫﺝ‬
‫ﺍﻟﺬﺍﺗﻴﺔ ﺍﻻﳓﺪﺍﺭ ﻣﻦ ﺩﺭﺟﺔ ﺑﺴﻴﻄﺔ‪ ،‬ﻟﻌﻠﻨﺎ ﻧﺴﺘﻨﺘﺞ ﻗﺎﻋﺪﺓ ﻋﺎﻣﺔ ﻟﻠﺘﻌﺮﻑ ﻋﻠﻰ ﻫﺬﺍ‬
‫ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ‪ ،‬ﻣﻦ ﺧﻼﻝ ﺩﺭﺍﺳﺔ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺃﻭ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬
‫ﺍﳉﺰﺋﻴﺔ ﻭﺍﻟﱵ ﻳﺄﰐ ﺷﺮﺣﻬﺎ ﻓﻴﻤﺎ ﺑﻌﺪ‪.‬‬
‫ﳓﺎﻭﻝ ﺍﻵﻥ ﺍﻟﺒﺤﺚ ﰲ ﺧﺼﺎﺋﺺ ﳕﻮﺫﺝ ﺍﳓﺪﺍﺭﻱ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﰒ‬
‫ﺍﻟﺜﺎﻧﻴﺔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y t = d + fy t - 1 + e t‬‬
‫ﻳﻌﺮﻑ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺑﺎﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‪ ،‬ﻭﻳﺮﻣﺰ ﻟﻪ ﺑـ‬
‫)‪ AR(1‬ﺍﻟﺬﻱ ﰎ ﺗﻨﺎﻭﻟﻪ ﻋﻨﺪ ﺩﺭﺍﺳﺔ ﻣﻮﺿﻮﻉ ﺍﻹﺳﺘﻘﺮﺍﺭﻳﺔ ﺃﻭ ﺍﻟﺴﻜﻮﻥ‪.‬‬
‫ﳝﻜﻦ ﻭﺑﺈﺩﺧﺎﻝ ﻣﻌﺎﻣﻞ ﺍﻟﺘﺄﺧﲑ )‪ (lag operator‬ﻛﺘﺎﺑﺔ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﻛﻤﺎ‬
‫ﻳﻠﻲ‪:‬‬
‫‪y t = d + f1 L yt + e t‬‬
‫‪Ly t = y t -1‬‬ ‫ﺣﻴﺚ‬
‫‪L2 y t = y t - 2‬‬ ‫ﻭ‬

‫‪156‬‬
 ‫ﻭﻫﻜﺬﺍ‪ ،‬ﻭﻣﻨﻪ‬
‫‪d‬‬ ‫‪et‬‬
‫= ‪yt‬‬ ‫‪+‬‬
‫) ‪(1 - f ) (1 - fL‬‬
‫‪d‬‬
‫= ‪E( yt ) = m‬‬
‫‪1-f‬‬
‫ﺣﱴ ﻳﻜﻮﻥ ﻝ ‪ m‬ﺣﻼﹰ ‪‬ﺎﺋﻴﺎ ﻳﺸﺘﺮﻁ ﺃﻥ ﺗﻜﻮﻥ ‪ ، f ¹ 1‬ﺑﻴﻨﻤﺎ ﺷﺮﻁ‬
‫ﺍﻻﺳﺘﻘﺮﺍﺭﻳﺔ ﻳﺘﻤﺜﻞ ﰲ ﺃﻥ ﺗﻜﻮﻥ ‪. f < 1‬‬
‫ﺑﺎﻓﺘﺮﺍﺽ ﺃﻥ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﻣﺴﺘﻘﺮ‪ ،‬ﻧﺒﺤﺚ ﺍﻵﻥ ﻋﻠﻰ ﺍﻟﺘﺒﺎﻳﻦ ‪ . g 0‬ﻭﻟﺘﺴﻬﻴﻞ‬
‫ﺍﻟﻌﻤﻠﻴﺔ ﻧﻔﺘﺮﺽ ﺃﻥ ‪. d = 0‬‬

‫‪g 0 = E ( y t ) 2 = E ( fy t - 1 + e t ) 2‬‬
‫‪= f 2 g 0 + s e2‬‬
‫‪s e2‬‬
‫=‬
‫‪1 - f2‬‬

‫ﺣﱴ ﻳﻜﻮﻥ ﻫﺬﺍ ﺍﳌﻘﺪﺍﺭ ﻣﻌﻘﻮﻻﹰ ﻛﺬﻟﻚ )‪‬ﺎﺋﻴﺎﹰ ﻭﻏﲑ ﺳﺎﻟﺐﹴ( ﻳ‪‬ﺸﺘﺮﻁ ﺃﻥ‬
‫ﺗﻜﻮﻥ ‪. f < 1‬‬
‫ﺗﻜﻮﻥ ﺍﻟﺘﺒﺎﻳﻨﺎﺕ ﺍﳌﺸﺘﺮﻛﺔ ﻣﻌﻄﺎﺓ ﺑﺎﻟﻌﻼﻗﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫] ‪g 1 = E ( y t y t - 1 ) = E [ ( fy t - 1 + e t ) y t - 1‬‬

‫) ‪= fE ( y t - 1 y t - 1 ) + E ( y t - 1 e t‬‬

‫‪= fg 0‬‬

‫ﻝ ‪y t -1‬‬‫ﺣﻴﺚ ‪ E ( y t -1 .e t ) = 0‬ﺍﻟﺬﻱ ﻳﺮﺟﻊ ﺇﱃ ﻛﻮﻥ ﺍﳋﻄﺄ ﺍﻟﻀﻤﲏ‬

‫ﻫﻮ ‪ ، e t - 1‬ﻭﺑﺎﻟﺘﺎﱄ ﻓﺈﻥ ‪ ، E ( e t - 1 . e t ) = 0‬ﻻﺧﺘﻼﻑ ﺍﻟﻔﺘﺮﺓ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬

‫‪157‬‬
 ‫] ‪g 2 = E ( y t y t - 2 ) = E [ ( fy t - 1 + e t ) y t - 2‬‬

‫‪= fg 1‬‬

‫‪= f ( fg 0 ) = f 2 g 0‬‬
‫‪g 3 = fg 2‬‬ ‫ﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ‬

‫‪= f( f 2g 0 ) = f 3g 0‬‬

‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ﻟﻜﻞ ‪ k = 1, K‬ﻭﺑﺎﻟﺘﺸﺎﺑﻪ‬

‫‪g k = fkg0‬‬

‫ﺗﺼﺒﺢ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬ ‫‪g0‬‬ ‫ﺑﻘﺴﻤﺔ ﻃﺮﰲ ﻫﺬﻩ ﺍﻟﻌﻼﻗﺔ ﻋﻠﻰ‬
‫ﳑﺜﻠﺔ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪rk = f k‬‬ ‫)‪(5.30‬‬

‫ﺣﻴﺚ ‪ ، r 0 = 1‬ﻭﺑﺴﺒﺐ ﺍﻟﻘﻴﺪ ﺍﳌﻔﺮﻭﺽ ﻋﻠﻰ ‪ f‬ﻣﻦ ﺃﺟﻞ ﺍﻹﺳﺘﻘﺮﺍﺭﻳﺔ‪،‬‬

‫ﺗﺒﺪﺃ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﰲ ﺍﻟﺘﻨﺎﻗﺺ ﻭﺍﻻﺿﻤﺤﻼﻝ ﻛﻤﺎ ﻫﻮ ﻣﺒﲔ ﰲ ﻫﺬﺍ ﺍﳌﺜﺎﻝ‪.‬‬

‫ﻣﺜﺎﻝ )‪ :(25‬ﺃﺭﺳﻢ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﻨﻤﻮﺫﺝ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪y t = 0. 8 y t - 1 + e t‬‬
‫ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻌﻼﻗﺔ ﺃﻋﻼﻩ‬
‫‪r 1 = f 1 = 0. 8‬‬
‫‪r 2 = f 2 = ( 0. 8 ) 2 = 0. 64‬‬
‫‪r 3 = 0. 512‬‬
‫‪r 4 = 0. 4096‬‬

‫‪158‬‬
‫‪r 10 = 0. 1074‬‬
‫‪r 15 = 0. 0352‬‬
‫‪...‬ﺍﱁ ‪.‬ﻭﻧﻌﱪ ﻋﻦ ﻫﺬﻩ ﺍﻷﺭﻗﺎﻡ ﺑﻴﺎﻧﻴﺎﹰ ﲟﺎ ﻳﻠﻲ‪:‬‬
‫)‪AR(1‬‬ ‫ﺍﻟﺸﻜﻞ )‪ : (27‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻨﻤﻮﺫﺝ‬

‫ﺑﻴﻨﻤﺎ ﺇﺫﺍ ﻛﺎﻧﺖ ‪ f = -0. 8‬ﻓﺈﻥ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺗﺄﺧﺬ ﺍﻟﺸﻜﻞ‬

‫)‪AR(1‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(28‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻨﻤﻮﺫﺝ‬

‫‪159‬‬
‫ﻋﻜﺲ ﻣﺎ ﺳﺒﻖ‪ ،‬ﻳﺼﻌﺐ ﲢﺪﻳﺪ ﺩﺭﺟﺔ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻻﳓﺪﺍﺭﻱ ﻣﻦ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﻟﱵ‬
‫ﺗﻮﻓﺮﻫﺎ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‪ ،‬ﻛﻮ‪‬ﺎ ﺗﺒﻘﻰ ﻣﺴﺘﻤﺮﺓ ﺍﻟﺘﺪﻫﻮﺭ)ﻣﻀﻤﺤﻠﺔ( ﰲ ﺣﺎﻟﺔ‬
‫ﺍﻻﺳﺘﻘﺮﺍﺭ ﻭﻻ ﺗﻨﻌﺪﻡ ﺑﺴﺮﻋﺔ؛ ﻭﻟﺘﻮﺿﻴﺢ ﺃﻛﺜﺮ ﻧﺪﺭﺝ ﺍﳌﺜﺎﻝ ﺍﻟﺘﺎﱄ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻟﺜﺎﻧﻴﺔ‬
‫)‪.AR(2‬‬

‫‪y t = d + f 1 y t -1 + f 2 y t - 2 + e t‬‬
‫‪y t = d + f 1 Ly t + f 2 L 2 y t + e t‬‬
‫‪d‬‬ ‫‪et‬‬
‫= ‪yt‬‬ ‫‪+‬‬
‫) ‪1 - ( f 1 + f 2 ) ( 1 - fL - f 2 L 2‬‬
‫‪d‬‬
‫= ‪E( y t ) = m‬‬
‫) ‪1 - ( f1 + f 2‬‬
‫ﺍﳋﺎﺻﺔ )‪ (Characteristic roots‬ﳍﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﳝﻜﻦ‬
‫‪،3‬‬
‫ﺑﺪﺭﺍﺳﺔ ﺍﳉﺬﻭﺭ‬
‫ﺍﺳﺘﻨﺘﺎﺝ ﺍﻟﺸﺮﻁ ﺍﻟﻀﺮﻭﺭﻱ ﻟﻼﺳﺘﻘﺮﺍﺭ ﻭﺍﳌﺘﻤﺜﻞ ﰲ ‪:‬‬
‫‪4‬‬

‫‪f1 + f 2 < 1‬‬

‫)‪AR(P‬‬ ‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ﻭﰲ ﳕﻮﺫﺝ‬
‫‪f 1 + f 2 + f 3 +... + f p < 1‬‬
‫ﺗﺘﻤﺜﻞ ﺗﺒﺎﻳﻨﺎﺕ ﳕﻮﺫﺝ )‪ AR(2‬ﰲ‬
‫] ‪g 0 = E ( y t ) 2 = E [ ( f 1 y t -1 + f 2 y t - 2 + e t ) y t‬‬

‫‪= f 1 g 1 + f 2 g 2 + s e2‬‬

‫‪ - 3‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ ‪:‬‬

‫‪1-A.C Harvey ,Time-Series Models (1981) , P32.‬‬
‫‪2- R.S.Pindyck and D.L.Rubinfeld, Econometric Models and Economic Forecast, 1981,P.‬‬
‫‪35.‬‬
‫‪ - 4‬ﻭﻣﻦ ﺍﻟﺸﺮﻭﻁ ﺍﻷﺧﺮﻯ ﳚﺐ ﲢﻘﻖ ﻣﺎ ﻳﻠﻲ ﺃﻳﻀﺎ‪:‬‬
‫‪f 2 - f1 < 1‬‬
‫‪f2 < 1‬‬
‫‪f1 < 2‬‬

‫‪160‬‬
 g 1 = E [ ( f 1 y t -1 + f 2 y t -2 + e t ) y t -1 ]

= f1g 0 + f2g 1

g 2 = E [ ( f 1 y t -1 + f 2 y t - 2 + e t ) y t -2 ]

= f1g 1 + f 2g 2
‫ﻣﻨﻪ ﻓﺈﻥ‬
g 3 = f1g 2 + f 2g 1

g k = f1g k -1 + f2g k-2 ‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬

‫ ﳓﺼﻞ ﻋﻠﻰ‬، g 0 ‫ﺑﺘﻘﺴﻴﻢ ﺍﳌﻌﺎﺩﻻﺕ ﺍﻟﺴﺎﺑﻘﺔ ﻭﻛﺎﻟﻌﺎﺩﺓ ﻋﻠﻰ‬

g 1 f1g 0 + f 2g 1
r1 = =
g0 g0
f1
=
1 - f2

g 2 f1 g 1 + f 2 g 0
r2 = =
g0 g0

r 2 = f 1r1 + f 2
f1
= f1 ( ) + f2
1 - f2
f 12
=( ) + f2
1 - f2

r 3 = f 1r 2 + f 2r 1

‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬

161
 ‫‪r k = f 1r k -1 + f 2r k -2‬‬

‫ﺑﻴﻨﻤﺎ ﰲ ﳕﻮﺫﺝ ﻣﻦ ﺍﻟﺪﺭﺟﺔ )‪ ،(p‬ﻓﺈﻥ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻟﺔ ﺗﺄﺧﺬ ﺍﻟﺸﻜﻞ‬

‫‪r k = f 1 r k - 1 + f 2 r k - 2 +... + f p r k - p‬‬ ‫)‪(5.31‬‬

‫ﺍﻟﱵ ﻳﻄﻠﻖ ﻋﻠﻴﻬﺎ ﲟﻌﺎﺩﻻﺕ ﻳﻮﻝ ﻭﻟﻜﺮ ‪ Yule-Walker equations‬ﺍﻟﱵ‬

‫ﺳﻨﺴﺘﻌﺮﺽ ﺍﺳﺘﺨﺪﺍﻣﺎ‪‬ﺎ ﻓﻴﻤﺎ ﺑﻌﺪ‪.‬‬
‫ﺇﻥ ﺍﻟﻨﺘﻴﺠﺔ ﺍﳌﻤﻜﻦ ﺍﻟﺘﻮﺻﻞ ﺇﻟﻴﻬﺎ ﻣﻦ ﻫﺬﻩ ﺍﻷﺷﻜﺎﻝ ﺍﳌﺨﺘﻠﻔﺔ ﺑﺎﻟﻨﺴﺒﺔ‬
‫ﻟﻠﻨﻤﺎﺫﺝ ﺍﳌﻌﺮﻭﺿﺔ‪ ،‬ﺃﻥ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺗﻨﻄﻠﻖ ﻣﻦ ﺍﻟﻮﺍﺣﺪ ﻭﺗﺒﻘﻰ‬
‫ﻣﺴﺘﻤﺮﺓ ﺍﻟﺘﺪﻫﻮﺭ ﺃﻭ ﺍﻻﺿﻤﺤﻼﻝ‪ ،‬ﺑﻴﻨﻤﺎ ﺗﺒﺘﺮ ﰲ ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‬
‫ﻣﺒﺎﺷﺮﺓ ﻋﻨﺪ ﺍﻟﺪﺭﺟﺔ ) ‪ . ( q‬ﰲ ﺣﺎﻟﺔ )‪ MA(2‬ﻣﺜﻼ ﺗﺒﺘﺮ ﻫﺬﻩ ﺍﳌﻌﺎﻣﻼﺕ ﻋﻨﺪ‬
‫ﺍﻟﺪﺭﺟﺔ ﺍﻟﺜﺎﻧﻴﺔ ﻭﻫﻜﺬﺍ‪ .‬ﳍﺬﺍ ﻻ ﳝﻜﻦ ﺍﻻﺳﺘﻌﺎﻧﺔ ‪‬ﺬﻩ ﺍﻟﺪﺍﻟﺔ ﻟﺘﺤﺪﻳﺪ ﺩﺭﺟﺔ ﳕﺎﺫﺝ‬
‫ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ‪ ،‬ﻟﻜﻨﻪ ﻳﺴﺘﻌﺎﻥ ‪‬ﺎ ﻟﻸﻏﺮﺍﺽ ﺍﻷﺧﺮﻯ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ (1‬ﺗﻘﻴﺲ ﺩﺭﺟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺑﲔ ﻣﺸﺎﻫﺪﺍﺕ ﺍﻟﻈﺎﻫﺮﺓ ﺍﳌﺪﺭﻭﺳﺔ‪.‬‬
‫‪ (2‬ﺗﺴﺎﻋﺪ ﰲ ﲢﺪﻳﺪ ﺩﺭﺟﺔ ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‪.‬‬
‫‪ (3‬ﺗﻌﻜﺲ ﻣﺪﻯ ﺇﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﺍﻟﺬﻱ ﻳﺘﺠﻠﻰ ﰲ ﺃﻥ‬
‫ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺗﺘﻼﺷﻰ ﺑﺴﺮﻋﺔ ﺃﻱ ﻗﺒﻞ ﺍﻟﺪﺭﺟﺔ ‪ K‬ﻭﺍﻟﱵ‬
‫ﺗﻌﺎﺩﻝ) ‪.( T‬‬
‫‪4‬‬
‫‪ (4‬ﻛﺸﻒ ﺃﺳﺒﺎﺏ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﻣﻦ ﺍﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﻓﺼﻠﻴﺔ‪...،‬ﺍﱁ ‪.‬‬

‫‪Partial Autocorrelation‬‬ ‫ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ‪:‬‬

‫ﺃﻣﺎﻡ ﺍﻟﻮﺿﻊ ﺍﻟﺼﻌﺐ ﺍﻟﺬﻱ ﻋﺮﻓﻨﺎﻩ ﻟﻠﺘﻌﺮﻑ ﻋﻠﻰ ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ‬
‫)‪ ،AR(p‬ﺣﱴ ﻭﺇﻥ ﻛﺎﻥ ﻣﻦ ﺩﺭﺟﺔ ﺑﺴﻴﻄﺔ ﻣﻦ ﺧﻼﻝ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‪ ،‬ﻧﻮﺩ‬
‫ﺍﻻﺳﺘﻌﺎﻧﺔ ﺑﺪﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ ﻭﺍﻟﱵ ﺳﻨﺸﺮﺣﻬﺎ ﰲ ﻫﺬﻩ ﺍﻟﺴﻄﻮﺭ ﺍﻟﻘﻠﻴﻠﺔ‪.‬‬

‫‪162‬‬
‫ﺇﻥ ﺷﻜﻞ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺬﻱ ﻧﺮﻳﺪ ﲢﺪﻳﺪﻩ‪ ،‬ﳜﻀﻊ ﻟﻠﺸﻜﻞ ﺍﻟﻌﺎﻡ ﺍﻟﺘﺎﱄ‬
‫)‪:AR(p‬‬
‫‪y t = f 1 y t - 1 + f 2 y t - 2 + ... + f p y t - p + e t‬‬

‫ﻣﻘﺎﺑﻞ ‪. k‬‬ ‫‪f$ k‬‬ ‫ﳝﻜﻦ ﺗﻌﺮﻳﻒ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺑﺄ‪‬ﺎ ﲤﺜﻴﻞ ﺑﻴﺎﱐ ﳌﻌﺎﻣﻼﺕ ﺍﻟﺪﺍﻟﺔ‬
‫ﻟﺘﻮﺿﻴﺢ ﻛﻴﻔﻴﺔ ﺣﺴﺎﺏ ﻫﺬﻩ ﺍﳌﻌﺎﻣﻼﺕ ﻧﺪﺭﺝ ﺍﻟﻄﺮﻳﻘﺘﲔ ﺍﻟﺘﺎﻟﻴﺘﲔ‬

‫ﺍ ‪ -‬ﺍﻟﻄﺮﻳﻘﺔ ﺍﻻﳓﺪﺍﺭﻳﺔ‪:‬‬
‫ﺗﺘﻤﺜﻞ ﻫﺬﻩ ﺍﻟﺘﻘﻨﻴﺔ ﰲ ﲢﺪﻳﺪ ﺃﻭﻻ ﺍﻟﺪﺭﺟﺔ ‪ ، k = T‬ﰒ ﺇﺟﺮﺍﺀ ﻋﻤﻠﻴﺔ‬
‫‪4‬‬
‫ﺗﻘﺪﻳﺮ ﻟـ ‪ y t‬ﻋﻠﻰ ‪ y‬ﻭﺍﳊﺼﻮﻝ ﻋﻠﻰ ‪ ، f$ 1‬ﰒ ‪ y t‬ﻋﻠﻰ ‪ y t - 2‬ﻭﺍﳊﺼﻮﻝ‬ ‫‪t -1‬‬

‫ﻋﻠﻰ ‪ f$ 2‬ﻭﻫﻜﺬﺍ ﺇﱃ ﻏﺎﻳﺔ ﺍﳊﺼﻮﻝ ﻋﻠﻰ ‪ ، f$ k‬ﻭﺗﺘﺤﺪﺩ ﺍﻟﺪﺭﺟﺔ ‪ p‬ﳌﺎ ﺗﻨﻌﺪﻡ ﺃﻭ‬
‫ﺗﻘﺘﺮﺏ ﻣﻨﻪ ‪ ، f$ k‬ﺣﻴﺚ ‪. k > p‬‬

‫ﻳﻮﻝ_ﻭﻟﻜﺮ‪Yule-Walker :‬‬ ‫ﺏ‪ -‬ﻃﺮﻳﻘﺔ ﻣﻌﺎﺩﻻﺕ‬

‫ﻭﺳﻨﺮﺍﻫﺎ ﺍﻵﻥ ﺃﻱ‬ ‫)‪AR(2‬‬ ‫ﻭ‬ ‫)‪AR(1‬‬ ‫ﺭﺃﻳﻨﺎ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻻﺕ ﰲ ﺣﺎﻟﺔ‬
‫ﻣﻌﺎﺩﻻﺕ ﻳﻮﻝ‪-‬ﻭﻟﻜﺮ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y t = f 1 y t - 1 + f 2 y t - 2 + ... + f p y t - p + e t‬‬

‫ﺑﻀﺮﺏ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻟﺔ ﰲ ‪ y t‬ﻭ ‪ y t - 1‬ﻭ ‪ ...‬ﻭ ‪ y t - p‬ﻭ ‪ ...‬ﻭ ‪ y t - k‬ﰒ‬

‫ﺃﺧﺬ ﺍﻟﺘﻮﻗﻊ ﺍﻟﺮﻳﺎﺿﻲ ﳍﺎ‪ ،‬ﳓﺼﻞ ﻋﻠﻰ ﺍﻟﺘﺒﺎﻳﻨﺎﺕ ﻭﺍﻟﺘﺒﺎﻳﻨﺎﺕ ﺍﳌﺸﺘﺮﻛﺔ ‪، g 1 ، g 0‬‬
‫‪ g k ،...، g p ،... ، g 2‬ﻭﻛﺎﻵﰐ‪:‬‬

‫‪g 0 = E ( y t ) 2 = f 1 g 1 + f 2 g 2 +... + f p g p + s e2‬‬

‫‪163‬‬
 ‫‪g 1 = E ( y t y t - 1 ) = f 1 g 0 + f 2 g 1 +... + f p g p- 1‬‬

‫‪g 2 = E ( y t y t - 2 ) = f 1 g 1 + f 2 g 0 + f 3 g 1 +... + f p g p - 2‬‬

‫‪g p = E ( y t y t - p ) = f 1 g p-1 + f 2 g p- 2 +... + f p g 0‬‬

‫ﻛﻤﺎ ﺭﺃﻳﻨﺎ ﰲ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﰲ ﺣﺎﻟﺔ )‪ AR(1‬ﻭ)‪ ،AR(2‬ﻓﺈ‪‬ﺎ ﻻ‬

‫ﺗﺒﺘﺮ ﻣﺒﺎﺷﺮﺓ ﻋﻨﺪ ﺍﻟﺪﺭﺟﺔ ‪ ، p‬ﺑﻞ ﻳﺴﺘﻤﺮ ﻫﺬﺍ ﺍﻻﺭﺗﺒﺎﻁ ﺇﱃ ﺃﺟﻞ ﻣﺴﻤﻰ ﻭﻏﲑ‬
‫ﻣﻌﺮﻭﻑ ﳊﺪ ﺍﻵﻥ‪ ،‬ﻭﳍﺬﺍ ﻓﺈﻥ ﺍﻟﺘﺒﺎﻳﻨﺎﺕ ﺍﳌﺸﺘﺮﻛﺔ ﻻ ﺗﻨﺘﻬﻲ ﻋﻨﺪ ﻫﺬﻩ ﺍﻟﺪﺭﺟﺔ ﺑﻞ‬
‫ﺗﺒﻘﻰ ﻣﺴﺘﻤﺮﺓ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪g‬‬ ‫‪k‬‬ ‫‪= E ( y t y t -1 ) = f 1 g‬‬ ‫‪k -1‬‬ ‫‪+ f2g‬‬ ‫‪k -2‬‬ ‫‪+ ... + f p g‬‬ ‫‪k-p‬‬

‫ﺑﻘﺴﻤﺔ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻻﺕ ﺍﳌﺨﺘﻠﻔﺔ ﻋﻠﻰ ﺍﻟﺘﺒﺎﻳﻦ‪ ،‬ﳓﺼﻞ ﻋﻠﻰ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ‬
‫ﺍﻻﺭﺗﺒﺎﻁ ﻫﺬﻩ ﰲ ﺷﻜﻞ ﻣﻌﺎﺩﻻﺕ ﻳﻮﻝ‪-‬ﻭﻟﻜﺮ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪r 1 = f 1 + f 2 r 1 +... + f pr p- 1‬‬
‫‪r 2 = f 1 r 1 + f 2 +... + f pr p- 2‬‬
‫‪...‬‬

‫‪r p = f 1 r p- 1 + f 2 r p- 2 +... + f p‬‬

‫‪...‬‬

‫‪r k = f 1 r k -1 + f 2 r k - 2 + .... + f p r k - p‬‬

‫ﻭﻫﻲ ﻣﻌﺎﺩﻻﺕ ﻳﻮﻝ‪-‬ﻭﻟﻜﺮ‪.‬‬
‫ﻓﺒﻤﻌﺮﻓﺔ ﻣﻌﻠﻤﺎﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ‪ r k‬ﺣﻴﺚ ) ‪ ( k = 1, K‬ﳝﻜﻦ‬
‫ﻣﻌﺮﻓﺔ ‪ f p ،...، f 1‬ﻭﺍﻟﻌﻜﺲ ﺻﺤﻴﺢ‪ ،‬ﻭ ﳍﺬﺍ ﳝﻜﻦ ﺍﳊﺼﻮﻝ ﻋﻠﻰ ﻫﺬﻩ ﺍﻷﺧﲑﺓ‬
‫ﺑﻌﺪ ﺗﻌﻮﻳﺾ ‪ r k‬ﲟﻘﺪﺭﺍ‪‬ﺎ ﺃﻱ ‪ rk‬ﻣﻦ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﻌﻴﻨﺔ ‪ ،‬ﻭﻟﻜﻦ‬
‫‪5‬‬

‫ﺍﳌﺸﻜﻞ ﺍﻟﺬﻱ ﻳﺒﻘﻰ ﻣﻄﺮﻭﺣﺎ ﺑﺈﳊﺎﺡ ﻫﻮ ﻛﻴﻔﻴﺔ ﲢﺪﻳﺪ ﺍﻟﺪﺭﺟﺔ ) ‪ ( p‬ﻭﳍﺬﺍ ﺳﻨﻌﻮﺩ‬

‫) ‪(128‬‬ ‫‪ - 5‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ ﺍﻟﺼﻔﺤﺔ‬

‫‪164‬‬
‫ﰒ ﺍﻓﺘﺮﺍﺽ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ‬ ‫‪rk‬‬ ‫ﺇﱃ ﻣﻌﺎﺩﻻﺕ ﻳﻮﻝ‪-‬ﻭﻟﻜﺮ ﻭﺗﻌﻮﻳﺾ ‪ r k‬ﲟﻘﺪﺭﺍ‪‬ﺎ‬
‫ﲣﻀﻊ ﻟﻨﻤﻮﺫﺝ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﰒ ﺍﻟﺜﺎﻧﻴﺔ ﻭﻫﻜﺬﺍ‪.‬‬
‫^‬ ‫^‬
‫‪p = 1 : AR ( 1 ) Þ r 1 = r1 = f 1‬‬

‫ﰒ‬
‫^ ‪ì‬‬ ‫^‬ ‫^‬
‫‪ü‬‬
‫‪ïï‬‬ ‫‪r‬‬ ‫‪1‬‬ ‫=‬ ‫‪r‬‬‫‪1‬‬ ‫=‬ ‫‪f‬‬ ‫‪1‬‬ ‫‪+‬‬ ‫‪f‬‬ ‫‪2 r1 ï‬‬
‫‪ï‬‬
‫‪p = 2 : AR ( 2) Þ í‬‬ ‫‪ý‬‬
‫^ ‪ï‬‬ ‫^‬ ‫‪^ ï‬‬

‫‪ïî r 2 = r2 = f 1 r1 + f 2 ïþ‬‬
‫ﻭﺑﺎﻟﺘﻌﻮﻳﺾ ﳓﺼﻞ ﻋﻠﻰ ‪ f$ 1‬ﻭ ‪ f$ 2‬ﻭﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬

‫‪f$ 1 = a 1‬‬ ‫ﳓﺼﻞ ﻋﻠﻰ‬ ‫‪p=1‬‬ ‫ﳌﺎ‬

‫‪f$ 2 = a 2‬‬ ‫‪ f$ 1‬ﻭ ‪ f$ 2‬ﻭﻧﺄﺧﺬ‬ ‫ﳓﺼﻞ ﻋﻠﻰ‬ ‫‪p= 2‬‬ ‫ﳌﺎ‬
‫‪f$ 3 = a 3‬‬ ‫‪ f$ 1‬ﻭ ‪ f$ 2‬ﻭ ‪ f$ 3‬ﻭﻧﺄﺧﺬ‬ ‫ﳓﺼﻞ ﻋﻠﻰ‬ ‫‪p= 3‬‬ ‫ﳌﺎ‬
‫‪ . a k‬ﻭﳝﻜﻦ ﺍﻟﺘﺄﻛﺪ ﻣﻦ ﻫﺬﺍ‬ ‫‪= 0‬‬ ‫ﻓﺈﻧﻨﺎ ﻧﺘﻮﻗﻊ ﺃﻥ ﻳﻜﻮﻥ‬ ‫‪k >p‬‬ ‫ﳌﺎ‬
‫ﺑﺎﺳﺘﺨﺪﺍﻡ ﺇﺣﺼﺎﺀﺓ ﺳﺘﻴﻮﺩﻧﺖ ﻻﺧﺘﺒﺎﺭ ﺍﻟﻔﺮﺿﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ ﻋﻨﺪ ﻛﻞ ﻣﺮﺣﻠﺔ‪:‬‬
‫‪H a:a k ¹ 0‬‬ ‫ﻣﻘﺎﺑﻞ‬ ‫‪H 0:a k = 0‬‬
‫‪a$ k - a k‬‬
‫=‪t cal‬‬ ‫) ‪® N( 0, 1‬‬
‫) ‪se( a$ k‬‬

‫‪ ، se( $a‬ﻛﻤﺎ ﳝﻜﻦ ﺍﺳﺘﺨﺪﺍﻡ ﺍﺧﺘﺒﺎﺭ ﺑﺎﺭﺗﻠﺖ‬ ‫‪k‬‬ ‫=)‬

‫‪1‬‬ ‫ﺃﻳﻦ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ‬
‫‪T‬‬
‫ﺍﻟﺬﻱ ﺳﻨﺘﻨﺎﻭﻟﻪ ﻓﻴﻤﺎ ﺑﻌﺪ‪.‬‬

‫‪165‬‬
‫ﻧﺴﻤﻲ ﻫﺬﻩ ﺍﳌﻘﺪﺭﺍﺕ ﲟﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ‪ ،‬ﺍﻟﱵ ﺗﺒﺘﺮ‬
‫ﻣﺒﺎﺷﺮﺓ ﺑﻌﺪ ﺍﻟﺪﺭﺟﺔ) ‪.( p‬‬
‫)‪AR(2‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(29‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ ﻟﻨﻤﻮﺫﺝ‬

‫ﻣﻦ ﺧﻼﻝ ﻣﺎ ﺳﺒﻖ‪ ،‬ﺭﺃﻳﻨﺎ ﺃﻥ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﰲ ﺣﺎﻟﺔ )‪ MA(q‬ﻭﺩﺍﻟﺔ‬

‫ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ ﰲ ﺣﺎﻟﺔ )‪ AR(p‬ﺗﺒﺘﺮ ﻣﺒﺎﺷﺮﺓ ﺑﻌﺪ ﺍﻟﺪﺭﺟﺔ ‪ q‬ﻭ ‪ p‬ﻋﻠﻰ‬
‫ﺍﻟﺘﺮﺗﻴﺐ ﻧﻈﺮﻳﺎ‪ ،‬ﺇﻻ ﺃﻧﻪ ﻭﰲ ﺍﻟﻮﺍﻗﻊ ﺍﻟﻌﻤﻠﻲ‪ ،‬ﻻ ﻳﺘﻢ ﺫﻟﻚ ‪‬ﺬﻩ ﺍﻟﺴﻬﻮﻟﺔ‪ ،‬ﻭﳍﺬﺍ ﻭﺟﺐ‬
‫ﺍﺳﺘﻌﻤﺎﻝ ﺃﺩﻭﺍﺕ ﺇﺧﺘﺒﺎﺭﻳﺔ ﺧﺎﺻﺔ ﻟﻠﺒﺤﺚ ﰲ ﻣﻌﻨﻮﻳﺔ ﻫﺬﻩ ﺍﳌﻌﺎﻣﻼﺕ ﰒ ﲢﺪﻳﺪ ﰒ‬
‫ﻋﺰﻝ ﺍﳌﻌﻠﻤﺎﺕ ﻏﲑ ﺍﻟﺼﻔﺮﻳﺔ ﻟﻜﻞ ﺩﺍﻟﺔ‪ ،‬ﰒ ﲢﺪﻳﺪ ﺩﺭﺟﺔ ﺍﻟﻨﻤﻮﺫﺝ ﺇﻥ ﺃﻣﻜﻦ‪.‬‬

‫‪ -3-1.2‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ‬
‫‪6‬‬
‫)‪ARMA(p,q‬‬

‫ﺗﺸﻤﻞ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ‪-‬ﻛﻤﺎ ﻳﻈﻬﺮ ﰲ ﺍﻟﻜﺘﺎﺑﺔ ﺍﻟﻼﺗﻴﻨﻴﺔ ﺃﻋﻼﻩ‪ -‬ﻋﻠﻰ ﺍﻟﻘﺴﻢ‬
‫ﺍﻻﳓﺪﺍﺭﻱ ﺫﻱ ﺍﻟﺪﺭﺟﺔ) ‪ ( p‬ﻭﻗﺴﻢ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ) ‪ ( q‬ﺍﻟﱵ‬
‫ﺗﻜﺘﺐ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﻌﺎﻡ ﺍﻟﺘﺎﱄ‪.:‬‬

‫‪6- Autoregressive Moving Average of order p and q .‬‬

‫‪166‬‬
 ‫‪y t = f1 yt-1 +f2 y t-2 +... +fpy t-p +et +q1et-1 +... +qq et-q‬‬ ‫)‪(5.32‬‬

‫ﻭﺑﺈﺩﺧﺎﻝ ﻣﻌﺎﻣﻞ ﺍﻟﺘﺄﺧﲑ ﻓﺈﻥ ‪،‬‬

‫‪y t = f1 Ly t + f2 L2 y t +... +f p Lp y t + e t + q1 Le t +... +qq Lq e t‬‬

‫‪y t - f1 Ly t - f 2 L2 y t -... -f p Lp y t = e t + q1 Le t +... +qq Lq e t‬‬

‫‪( 1 - f 1 L - f 2 L 2 -... - f p L p ) y t = ( 1 + q 1 L +... + q q L q ) e t‬‬

‫‪f ( L) yt = q (L )e t‬‬
‫ﻛﺎﻟﻌﺎﺩﺓ ﻭﻻﺳﺘﻨﺒﺎﻁ ﺍﳋﺼﺎﺋﺺ ﺍﳉﻮﻫﺮﻳﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﱵ ﲣﻀﻊ ﳍﺬﺍ‬
‫ﺍﻟﻨﻮﻉ‪ ،‬ﻧﺪﺭﺝ ﺍﳌﺜﺎﻝ ﺍﻟﺘﺎﱄ ﺍﻟﺬﻱ ﻧﻔﺘﺮﺽ ﻓﻴﻪ ﺃﻧﻪ ﻟﺪﻳﻨﺎ ﳕﻮﺝ )‪ ARMA(1,1‬ﺍﻟﺬﻱ‬
‫ﻳ‪‬ﻜﺘﺐ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y t = f 1 y t -1 + e t + q 1 e t -1‬‬

‫ﻟﻠﺤﺼﻮﻝ ﻋﻠﻰ ﺍﻟﺘﺒﺎﻳﻨﺎﺕ ﺍﳌﺸﺘﺮﻛﺔ ﳓﺴﺐ ﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻟﺴﺎﺑﻘﺔ ﻋﻠﻰ ﻣﺎ‬
‫ﻳﻠﻲ‪:‬‬
‫} ) ‪g 0 = E{ y t ( f 1 y t -1 + e t + q 1 e t -1‬‬

‫}) ‪= E{( f1 y t -1 + e t + q1e t -1 )( f1 y t -1 + e t + q1 e t -1‬‬

‫‪= E ( f 1 y t -1 + e t + q 1 e t -1 ) 2‬‬

‫‪= f 12 g 0 + 2 f 1 q 1 + s e2 + q 12 s e2‬‬
‫) ‪s e2 ( 1 + 2 f 1 q 1 + q 12‬‬
‫= ‪g0‬‬
‫) ‪( 1 - f 12‬‬
‫ﻭﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ‬

‫‪167‬‬
 ‫) ‪s e2 ( f 1 + q 1 )( 1 + f 1 q 1‬‬
‫= ‪g1‬‬
‫) ‪( 1 - f 12‬‬

‫ﺃﻱ ﺃﻛﱪ ﻣﻦ ﺩﺭﺟﺔ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻓﺈﻥ‬ ‫‪k >1‬‬ ‫ﻭﳌﺎ ﺗﻜﻮﻥ‬

‫‪g 2 = f1g 1‬‬

‫‪g k = f 1 g k -1‬‬
‫ﻣﻨﻪ ﺗﻜﻮﻥ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬

‫) ‪( f 1 + q 1 )( 1 + f 1 q 1‬‬
‫= ‪r1‬‬
‫) ‪( 1 + f 1 q 1 + q 12‬‬

‫‪r k = f 1r k -1‬‬
‫‪.‬‬ ‫‪k >q =1‬‬ ‫ﳌﺎ ﺗﻜﻮﻥ‬

‫ﻧﻼﺣﻆ ﺑﺎﻟﻨﻈﺮ ﺇﱃ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺃﻋﻼﻩ‪ ،‬ﺃﻥ ﺳﻠﻮﻛﻬﺎ ﰲ‬

‫ﺣﺎﻟﺔ ‪ ، k > q = 1‬ﻳﻜﻮﻥ ﺍﳓﺪﺍﺭﻳﺎ ﲝﺘﺎﹰ‪ ،‬ﺃﻱ ﻣﺴﲑﺍﹰ ﻣﻦ ﺍﻟﻄﺮﻑ ﺍﻻﳓﺪﺍﺭﻱ ﻓﻘﻂ‪،‬‬
‫ﻭﻫﺬﺍ ﻷﻥ ﳕﻮﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ )‪ MA(1‬ﻻ ﺗﺘﺠﺎﻭﺯ ﺫﺍﻛﺮﺗﻪ ﻓﺘﺮﺓ ﻭﺍﺣﺪﺓ‪.‬‬
‫ﺇﻥ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻨﻤﻮﺫﺝ ﳐﺘﻠﻂ )‪ ARMA(1,1‬ﺗﻜﻮﻥ ﳌﺎ‪:‬‬

‫‪1.‬‬ ‫‪f 1 = 0. 3‬‬ ‫‪ q 1 = 0. 9.‬ﻭ‬

‫‪2.‬‬ ‫‪ q 1 = -0. 9 .‬ﻭ ‪f 1 = 0. 3‬‬

‫‪168‬‬
 ‫ﺍﻟﺸﻜﻞ)‪:(30‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻨﻤﻮﺫﺝ )‪ ARMA(1,1‬ﳌﺎ ‪f 1 = 0 . 3‬‬
‫ﻭ ‪q1 = 0.9‬‬

‫‪.q 1‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(31‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻨﻤﻮﺫﺝ )‪ ARMA(1,1‬ﳌﺎ ‪ f 1 = 0 . 3‬ﻭ ‪= - 0 . 9‬‬

‫ﳍﺬﺍ ﻧﺘﻮﻗﻊ ﺃﻥ ﺗﻜﻮﻥ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ ﻏﲑ ﻣﺴﻴ‪‬ﺮﺓ ﻣﻦ ﺍﻟﻄﺮﻑ‬

‫ﺍﻻﳓﺪﺍﺭﻱ ﺑﻌﺪ ﺍﻟﻔﺘﺮﺓ ‪ ، p‬ﺑﻞ ﻣﻮﺟﻬﺔ ﻣﻦ ﻃﺮﻑ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‪ ،‬ﳍﺬﺍ‬
‫ﻳﺼﻌﺐ ﺍﻟﺘﻌﺮﻑ ﻋﻠﻰ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ ﻭﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻮﺍﺭﺩﺓ ﺃﺩﻧﺎﻩ‪،‬‬
‫ﻛﻮﻥ ﺍﻟﺪﺍﻟﺘﲔ ﻣﺴﺘﻤﺮﰐ ﺍﻻﺿﻤﺤﻼﻝ )‪ ،(Dying out‬ﻓﻨﻌﺘﻤﺪ ﻋﻠﻰ ﺍﻟﺘﺠﺮﺑﺔ ﰒ‬
‫ﺍﳋﱪﺓ ﻓﻴﻤﺎ ﺑﻌﺪ ﻟﺘﺤﺪﻳﺪ ‪ p‬ﻭ ‪. q‬‬

‫‪169‬‬
 ‫ﺍﳌﺮﻛﺒﺔ‪ARIMA(p ,d ,q) :‬‬ ‫‪ -4.1.2‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ‬
‫ﻳﺴﻤﻰ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﺑﺎﻟﻨﻤﺎﺫﺝ ﺍﳌﺘﺠﺎﻧﺴﺔ ﻏﲑ ﺍﳌﺴﺘﻘﺮﺓ ﺃﻭ ﺍﳌﺨﺘﻠﻄﺔ ﺍﳌﺮﻛﺒﺔ‬
‫)‪ (Integrated‬ﻣﻦ ﺍﻟﺪﺭﺟﺔ ‪ ، d‬ﻭﻳﺮﻣﺰ ﺇﻟﻴﻬﺎ ﺑـ )‪ ،ARIMA(p,d,q‬ﲣﺘﻠﻒ ﻫﺬﻩ‬
‫‪7‬‬

‫ﻋﻦ ﺳﺎﺑﻘﺘﻬﺎ ﺧﺎﺻﺔ )‪ ARMA(p,q‬ﰲ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﳌﺪﺭﻭﺳﺔ ﺗﻜﻮﻥ ﻏﲑ‬

‫ﻣﺴﺘﻘﺮﺓ‪ .‬ﻹﺯﺍﻟﺔ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﳚﺐ ﺍﻟﺒﺤﺚ ﻋﻦ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻷﻧﺴﺐ ﻹﺑﻌﺎﺩ ﻣﺼﺪﺭ‬
‫ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ‪ .‬ﻧﻄﺒﻖ ﻃﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﳌﺮﺓ ﺃﻭ ﻣﺮﺗﲔ ﺇﺫﺍ ﻛﺎﻥ‬
‫ﻣﺼﺪﺭ‪ ‬ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﻫﻮ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﺑﻴﻨﻤﺎ ﻧﻄﺒﻖ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺩﺭﺟﺔ ﺍﻟﻔﺼﻠﻴﺔ‬
‫ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﺳﺎﺑﻘﺎﹰ ﻹﺯﺍﻟﺔ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ‪ .‬ﻳﺼﺒﺢ ﺍﻟﺮﻣﺰ ﺍﻟﻼﺗﻴﲏ ﻟﻠﻨﻤﻮﺫﺝ‬
‫‪8‬‬

‫)‪ . SARIMA( p, d , p)( P, D, Q‬ﻧﻄﺒﻖ ﻃﺮﻳﻘﺔ ﺑﻮﻛﺲ‪ -‬ﻛﻮﻛﺲ )‪(Box-Cox‬‬ ‫‪s‬‬

‫ﺍﳌﺘﻤﺜﻠﺔ ﰲ ﺷﻜﻠﻬﺎ ﺍﻟﺒﺴﻴﻂ ﰲ ﺍﻟﻠﻮﻏﺎﺭﻳﺘﻢ ﰲ ﺣﺎﻟﺔ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﺍﻟﻨﺎﺗﺞ ﻣﻦ‬

‫ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﰲ ﺍﻟﺘﺒﺎﻳﻦ‪ .‬ﻓﺈﺫﺍ ﺛﺒﺖ ﻟﺪﻳﻨﺎ ﻋﻦ ﻃﺮﻳﻖ ﺍﻟﻜﺸﻒ ﺍﻟﺒﻴﺎﱐ ﺃﻭ ﺍﻹﺣﺼﺎﺋﻲ‬
‫ﺃﻥ ﻣﺼﺪﺭ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ‪ y t‬ﻫﻮ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﻧﻄﺒﻖ ﺍﻟﻔﺮﻭﻕ ﻣﻦ‬
‫ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﳌﺮﺓ ﻓﺘﻜﻮﻥ ‪ ،d = 1‬ﻭﻧﻜﺘﺐ‬
‫‪w t = y t - y t -1‬‬

‫ﻓﺈﺫﺍ ﻛﺎﻧﺖ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻨﺎﲡﺔ ‪ w t‬ﻣﺴﺘﻘﺮﺓ‪ ،‬ﻳﻜﻮﻥ )‪ ARIMA (p,1,q‬ﻫﻮ‬

‫ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﺮﻏﻮﺏ‪ ،‬ﺑﻴﻨﻤﺎ ﺇﻥ ﱂ ﻳﺘﺤﻘﻖ ﺍﻻﺳﺘﻘﺮﺍﺭ ﻧﻄﺒﻖ ﺍﻟﻄﺮﻳﻘﺔ ﻧﻔﺴﻬﺎ ﻟﻠﻤﺮﺓ‬
‫ﺍﻟﺜﺎﻧﻴﺔ ﺃﻱ ‪ d = 2‬ﻭﻫﻜﺬﺍ ﺃﻱ‪:‬‬
‫‪z t = w t - w t -1‬‬

‫‪ - 7.‬ﺣﻴﺚ ‪ d‬ﲤﺜﻞ ﻋﺪﺩ ﻣﺮﺍﺕ ﺗﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﻟﻠﺤﺼﻮﻝ ﻋﻠﻰ ﺃﺧﺮﻯ ﻣﺴﺘﻘﺮﺓ‬
‫‪ -8‬ﺗﺸﲑ ‪ P,Q‬ﺇﱃ ﺩﺭﺟﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻴﻨﻤﺎ ‪ D‬ﺗﺮﻣﺰ ﺇﱃ ﻋﺪﺩ ﻣﺮﺍﺕ ﺗﻄﺒﻴﻖ ﺗﻘﻨﻴﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ‬
‫ﺍﻟﺪﺭﺟﺔ ‪ p‬ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ ﺑﻴﻨﻤﺎ ‪ s‬ﲤﺜﻞ ﺩﺭﺟﺔ ﺍﻟﻔﺼﻠﻴﺔ ‪ ،‬ﻓﺘﻜﻮﻥ ﻣﺜﻼﹰ ‪ 4‬ﰲ‬
‫ﺩﻭﺭﻳﺔ ﻣﻮﲰﻴﺔ‪.‬‬

‫‪170‬‬
 ‫)‪.ARIMA(p,2,q‬‬ ‫ﻟﻴﻜﻮﻥ ﺍﻟﻨﻤﻮﺫﺝ‬
‫ﻟﻠﺴﻠﺴﻠﺔ‬ ‫‪q‬‬ ‫ﻭ‬ ‫‪p‬‬ ‫ﻧﺘﺒﻊ ﻧﻔﺲ ﺃﺳﻠﻮﺏ ﺍﻟﺘﺤﺪﻳﺪ ﺍﻟﻮﺍﺭﺩ ﺃﻋﻼﻩ ﻟﺘﺤﺪﻳﺪ‬
‫ﺍﻟﻨﺎﲡﺔ‪.‬‬
‫ﻭﻗﺒﻞ ﺍﻟﺘﻄﺮﻕ ﺇﱃ ﻣﻮﺿﻮﻉ ﺍﻟﺘﻘﺪﻳﺮ‪ ،‬ﻧﻮﺩ ﺗﻠﺨﻴﺺ ﳎﻤﻞ ﺍﳋﻄﻮﺍﺕ‬
‫ﺍﻟﻀﺮﻭﺭﻳﺔ ﺃﺛﻨﺎﺀ ﺍﻟﻌﻤﻞ ﺍﻟﺘﻄﺒﻴﻘﻲ ﺍﳌﺘﻤﺜﻞ ﰲ ﺍﳌﺮﺍﺣﻞ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪ - 1‬ﺗﻜﻮﻥ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ )‪ (A.C‬ﻣﺆﺷﺮﺍﹼ ﻣﻬﻤﺎﹰ ﻟﻜﺸﻒ ﻋﺪﻡ‬

‫ﺇﺳﺘﻘﺮﺍﺭﻳﺔ ﺃﻭ ﻋﺪﻡ ﺳﻜﻮﻥ ﺳﻠﺴﻠﺔ ﺯﻣﻨﻴﺔ‪ .‬ﻓﻌﻨﺪﻣﺎ ﻻ ﺗﻨﻌﺪﻡ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺑﻌﺪ ﺍﻟﻔﺘﺮﺓ‬
‫‪) T‬ﺭﺑﻊ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ( ﺗﻜﻮﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻏﲑ ﻣﺴﺘﻘﺮﺓ ﺣﻴﺚ ﳚﺐ‬
‫‪4‬‬
‫ﺗﻄﺒﻴﻘﻴﺎ ﺃﻥ ﺗﻘﻊ ﻣﻌﺎﻣﻼﺕ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺧﺎﺭﺝ ﳎﺎﻝ ﺍﻟﺜﹼـﻘـﺔ ﺍﶈﺪﺩ ﺇﺣﺼﺎﺋﻴﺎﹰ‪ ،‬ﻭﻫﻨﺎ‬
‫ﻧﻜﻮﻥ ﺑﺼﺪﺩ ﺩﺭﺍﺳﺔ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺮﻛﺒﺔ ﻛﻤﺎ ﺗﻌﺘﱪ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻛﺎﺷﻒ ﻣﻬﻢ‬
‫ﻟﻠﻔﺼﻠﻴﺔ ﻣﻦ ﺧﻼﻝ ﺍﻟﻘﻤﻢ ﻭﺍﻟﻨﺘﻮﺀﺍﺕ ﺍﻟﱵ ﺗﻈﻬﺮ ﰲ ﺷﻜﻞ ﻣﻨﺘﻈﻢ ﻋﻠﻰ ﻫﺬﻩ‬
‫ﺍﻟﺪﺍﻟﺔ ‪.‬‬
‫‪9‬‬

‫‪ - 2‬ﺑﺎﻟﻨﺴﺒﺔ ﻟﻨﻤﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻣﻦ ﺍﻟﺪﺭﺟﺔ )‪ (q‬ﺗﺒﺘﺮ ﺩﺍﻟﺔ‬

‫ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻣﺒﺎﺷﺮﺓ ﺑﻌﺪ ﻫﺬﻩ ﺍﻟﺪﺭﺟﺔ‪ ،‬ﺑﻴﻨﻤﺎ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ ﺗﺒﻘﻰ‬
‫ﻣﺘﻨﺎﻗﺼﺔ ﺑﻌﺪ ﻫﺬﻩ ﺍﻟﻔﺘﺮﺓ ﻭﻟﻜﻨﻬﺎ ﻻ ﺗﻨﻌﺪﻡ ﳊﻈﻴﺎ‪.‬‬
‫‪ - 3‬ﺑﺎﻟﻨﺴﺒﺔ ﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﻣﻦ ﺍﻟﺪﺭﺟﺔ )‪ ،(p‬ﺗﺒﺘﺮ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ‬
‫ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ )‪ (P.A.C‬ﻣﺒﺎﺷﺮﺓ ﺑﻌﺪ ﺗﻠﻂ ﺍﻟﺪﺭﺟﺔ‪ ،‬ﺑﻴﻨﻤﺎ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬
‫ﺗﺒﻘﻲ ﻣﺘﻨﺎﻗﺼﺔ ﻭﻟﻜﻨﻬﺎ ﻻ ﺗﻨﻌﺪﻡ ﺑﻨﻔﺲ ﺍﻟﺴﺮﻋﺔ‪.‬‬
‫‪ - 4‬ﺃﻣﺎ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ‪ ،‬ﻓﺈﻥ ﺍﻟﺪﺍﻟﺘﲔ ﺗﺒﻘﻴﺎﻥ ﻣﺴﺘﻤﺮﰐ ﺍﻟﺘﺪﻫﻮﺭ‬
‫ﻟﻜﻨﻬﻤﺎ ﻻ ﺗﻨﻌﺪﻣﺎﻥ ﻋﻨﺪ ﺍﻟﺪﺭﺟﺘﲔ ﺍﳌﺬﻛﻮﺭﺗﲔ ﺳﺎﺑﻘﺎ‪.‬‬

‫‪ - 9‬ﺃﻧﻈﺮ ﰲ ﺫﻟﻚ ﺍﻟﺼﻔﺤﺔ )‪ (43‬ﺍﻟﺸﻜﻞ)‪ (11‬ﻣﻦ ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ‪.‬‬

‫‪171‬‬
 ‫ﺍﳉﺪﻭﻝ )‪ :(31‬ﺃﺷﻜﺎﻝ ﺩﻭﺍﻝ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬
‫ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ‬ ‫ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬ ‫ﻧﻮﻉ ﺍﻟﻨﻤﻮﺫﺝ‬
‫ﻏﲑ ﻣﻨﻌﺪﻣﺔ‪DIES OUT-‬‬ ‫ﺗﺒﺘﺮ ﺑﻌﺪ ﺍﻟﻔﺘﺮﺓ ‪q‬‬ ‫)‪MA(q‬‬
‫ﺗﺒﺘﺮ ﺑﻌﺪ ﺍﻟﻔﺘﺮﺓ ‪p‬‬ ‫ﻏﲑ ﻣﻨﻌﺪﻣﺔ‪DIES OUT-‬‬ ‫)‪AR(p‬‬
‫ﻏﲑ ﻣﻨﻌﺪﻣﺔ‪DIES OUT-‬‬ ‫ﻏﲑ ﻣﻨﻌﺪﻣﺔ‪DIES OUT-‬‬ ‫)‪ARMA(p,q‬‬

‫ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻔﺼﻠﻴﺔ‪:‬‬
‫ﺗﻜﻮﻥ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻏﲑ ﺍﻟﺴﻨﻮﻳﺔ ﰲ ﺍﻟﻐﺎﻟﺐ ﻣﺎ ﺗﻜﻮﻥ ﺫﺍﺕ ﻣﺮﻛﺒﺔ‬
‫ﺩﻭﺭﻳﺔ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ‪ p‬ﻛﻤﺎ ﺭﺃﻳﻨﺎ ﻣﻦ ﻗﺒﻞ‪ ،‬ﳍﺬﺍ ﻧﺮﻳﺪ ﺇﻋﻄﺎﺀ ﺻﻮﺭﺓ ﳍﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ‬
‫ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺬﻱ ﻗﺪ ﺗﺄﺧﺬ ﺩﺍﻟﺔ ﺍﺭﺗﺒﺎﻃﻪ ﺍﻟﺬﺍﺗﻴﺔ ﺍﻟﺸﻜﻞ )‪ (11‬ﺍﻟﻮﺍﺭﺩ ﰲ ﺍﻟﺼﻔﺤﺔ‬
‫)‪ .(45‬ﺗﺄﺧﺬ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻷﺷﻜﺎﻝ ﺍﳌﺨﺘﻠﻔﺔ ﺍﻟﺴﺎﺑﻘﺔ ﻣﻦ ) ‪، SMA ( Q‬‬
‫‪10‬‬
‫‪s‬‬

‫) ‪ ، SARMA ( P, Q ) ، SAR ( P‬ﺍﻟﱵ ﳝﻜﻦ ﲤﺜﻴﻠﻬﺎ ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ ﰲ ﺍﻷﺷﻜﺎﻝ‬ ‫‪s‬‬ ‫‪s‬‬

‫ﺍﻟﺘﺎﻟﻴﺔ ﳌﺎ ‪ s=12 ،P=1 ،Q=1‬ﲟﻌﲎ ﺃﻥ ﺍﻟﺪﻭﺭﻳﺔ ﺷﻬﺮﻳﺔ‪.‬‬

‫‪y t = f y t - 12 + e t‬‬

‫‪ SMA‬ﰲ ﺷﻜﻠﻪ ﺍﻟﺘﺎﱄ ‪:‬‬ ‫‪( 1 ) 12‬‬ ‫‪ ، SAR‬ﺑﻴﻨﻤﺎ ﻧﻜﺘﺐ‬ ‫‪( 1 ) 12‬‬ ‫ﻧﺮﻣﺰ ﻟﻪ ﺑـ‬
‫‪y t = qe t - 12 + e t‬‬

‫ﻳﻨﻤﺎ ﻧﻌﱪ ﻋﻦ ﳕﻮﺫﺝ ﻓﺼﻠﻲ ﳐﺘﻠﻂ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪y t = f y t - 12 + fe t - 12 + e t‬‬

‫‪ -10‬ﺣﻴﺚ ‪ P , Q‬ﺗﺪﻝ ﻋﻠﻰ ﺩﻭﺭﻳﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺍﳋﺎﺻﺔ ﺑﺎﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻭﺍﻟﻘﺴﻢ ﺍﻻﳓﺪﺍﺭﻱ‬
‫ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ ﺑﻴﻨﻤﺎ ‪ S‬ﺗﺪﻝ ﻋﻠﻰ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﻔﺼﻠﻲ ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﺳﺎﺑﻘﺎﹰ‪.‬‬

‫‪172‬‬
‫ﻭﳝﻜﻦ ﺗﻄﺒﻴﻘﻴﺎﹰ ﲢﺪﻳﺪ ﺍﻟﺪﺭﺟﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻟﻠﻨﻤﻮﺫﺝ ﻣﻦ ﺧﻼﻝ ﺍﻟﻨﻈﺮ ﰲ‬
‫ﺍﳌﻌﺎﻣﻼﺕ ﺍﻟﻔﺼﻠﻴﺔ ﻓﻘﻂ ﻟﻠﺪﺍﻟﺘﲔ )‪ (ac‬ﻭ )‪ (pac‬ﺍﳌﺬﻛﻮﺭﺗﲔ ﺳﺎﺑﻘﺎﹰ ﻭﺑﻨﻔﺲ‬
‫ﺍﳌﻨﻬﺠﻴﺔ ﺍﻟﺴﺎﺑﻘﺔ ﺃﻳﻀﺎﹰ‪.‬‬
‫)‪(62‬‬ ‫ﻟﺘﻮﺿﻴﺢ ﻛﻞ ﻣﺎ ﺳﺒﻖ ﻧﺪﺭﺝ ﺍﳌﺜﺎﻝ ﺍﻟﺘﺎﱄ ﺍﳋﺎﺹ ﺑﺴﻠﺴﻠﺔ ﺯﻣﻨﻴﺔ ﺫﺍﺕ‬
‫ﻣﺸﺎﻫﺪﺓ ﺧﺎﺻﺔ ﺑﺎﺳﺘﻬﻼﻙ ﻏﺎﺯ ﺍﻟﺒﻮﺗﺎﻥ ﺍﳌﻌﺒﺄ ﰲ ﻗﺎﺭﻭﺭﺍﺕ ‪ 13‬ﻛﻠﻎ ﻟﻠﻮﻻﻳﺎﺕ‬
‫ﺍﳋﻤﺴﺔ ﺍﻟﻮﺳﻄﻰ ﻟﻠﺠﺰﺍﺋﺮ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ‪ ،1992.02-1987.01‬ﻭﺍﻟﺸﻜﻞ )‪ (32‬ﳝﺜﻞ‬
‫ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﺴﻠﺴﻠﺔ ‪.GB13‬‬
‫‪11‬‬

‫ﻣﻦ ﺧﻼﻝ ﺍﻟﺘﻤﻌﻦ ﰲ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‪ ،‬ﻧﺴﺘﻨﺘﺞ ﻭﻋﻠﻰ ﻏﺮﺍﺭ ﻣﺎ ﻗﻠﻨﺎﻩ‬
‫ﺳﺎﺑﻘﺎ ﻣﺎ ﻳﻠﻲ‪:‬‬
‫ﺍﻟﺸﻜﻞ )‪ :(32‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟـ‪. T=62 ،GB13‬‬

‫‪ -1‬ﺗﻈﻬـﺮ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﳉﻮﻫﺮﻳﺔ ﻟﻠﺴﻠﺴﻠﺔ ﰲ ﺍﻟﻔﺘﺮﺓ ) ‪ ( T‬ﺍﻟﱵ ﺗﻌﺎﺩﻝ ﻫﻨﺎ‬

‫‪4‬‬
‫ﺗﻘﺮﻳﺒﺎ )‪.(16‬‬

‫‪Eviews‬‬ ‫‪ - 11‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﺑﺮﻧﺎﻣﺞ‬

‫‪173‬‬
‫‪ – 2‬ﺇﻥ ﺍﳌﺮﻛﺒﺔ ﺍﻟﺪﻭﺭﻳﺔ ﺑﺎﺭﺯﺓ ﻓﻴﻬﺎ‪ ،‬ﻟﺬﺍ ﻭﺟﺐ ﺃﺧﺬﻫﺎ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ ﻋﻨﺪ‬
‫ﺍﻟﻨﻤﺬﺟﺔ‪ ،‬ﺣﻴﺚ ﻧﺴﺠﻞ ﻗﻤ‪‬ﺔ ﻋﻨﺪ ﻛﻞ ‪ 12‬ﺷﻬﺮﺍﹼ ﻛﻮﻥ ﺍﳌﻌﻄﻴﺎﺕ ﺷﻬﺮﻳﺔ ‪.‬‬
‫‪ - 3‬ﺗﻌﺘﱪ ﻫﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮﺓ‪ ،‬ﻛﻮﻥ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﺭﺗﺒﺎﻃﻬﺎ ﺍﻟﺬﺍﺗﻴﺔ‬
‫ﺩﺧﻠﺖ ﳎﺎﻝ ﺍﻟﺜﻘﺔ ﺍﳌﺘﻤﺜﻞ ﰲ ﺍﳋﻂ ﺍﻟﻨﻘﻄﻲ ﻗﺒﻞ ﺍﻟﻔﺘﺮﺓ )‪ ،(16‬ﻭﺍﻟﺬﻱ ﺳﻨﺸﺮﺡ‬
‫ﻣﻌﻨﺎﻩ ﻭﻛﻴﻔﻴﺔ ﺍﺷﺘﻘﺎﻗﻪ ﺍﻵﻥ‪ ،‬ﺑﻐﻴﺔ ﺗﺴﻬﻴﻞ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻌﺮﻑ ﻋﻠﻰ ﻧﻮﻉ ﺍﻟﻨﻤﻮﺫﺝ‬
‫ﻭﺩﺭﺟﺘﻪ‪.‬‬
‫ﺗﻈﻬﺮ ﺃﳘﻴﺔ ﳎﺎﻝ ﺍﻟﺜﻘﺔ ﰲ ﺃﻥ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻻ ﺗﺒﺘﺮ ﻛﻤﺎ‬
‫ﺭﺃﻳﻨﺎ ﻧﻈﺮﻳﺎﹰ‪ ،‬ﺃﻱ ﺃﻥﹼ ﻣﻌﺎﻣﻼﺕ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﻻ ﺗﺼﺒﺢ ﻣﺴﺎﻭﻳﺔ ﻟﻠﺼﻔﺮ ﺑﻌﺪ ﺃﺟﻞ‬
‫ﻣﺴﻤﻰ‪.‬‬
‫ﻓﻴ‪‬ﻌﺘﱪ ﻛﻞ ﻣﻌﺎﻣﻞ ﺍﺭﺗﺒﺎﻁ ﺫﺍﰐ ﺳﻮﺍﺀً ﻛﺎﻥ ﰲ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻜﻠﻴﺔ‬
‫ﺃﻭ ﺍﳉﺰﺋﻴﺔ ﻣﻌﺪﻭﻣﺎﹰ ﺇﺫﺍ ﻭﻗﻊ ﺿﻤﻦ ﺍ‪‬ﺎﻝ ‪:‬‬

‫) ‪±2SE ( rk‬‬

‫‪1‬‬
‫‪1‬‬ ‫‪k -1‬‬
‫= ) ‪SE ( rk‬‬ ‫‪( 1 + 2 å ri2 ) 2‬‬ ‫ﺣﻴﺚ‬
‫‪T‬‬ ‫‪i=1‬‬

‫ﺍﻟﺬﻱ ﻳﻄﻠﻖ ﻋﻠﻴﻪ ﺑﺎﺧﺘﺒﺎﺭ ﺑﺎﺭﺗﻠﺖ ‪ BARTLETT test‬ﺍﻟﺬﻱ ﻳﺴﺘﻌﺎﻥ ﺑﻪ‬

‫ﰲ ﻋﺰﻝ ﺍﳌﻌﺎﻣﻼﺕ ﺍﻟﻀﻌﻴﻔﺔ ﺍﳌﻌﻨﻮﻳﺔ‪ ،‬ﺃﻱ ﺍﻟﻮﺍﻗﻌﺔ ﺩﺍﺧﻞ ﻫﺬﺍ ﺍ‪‬ﺎﻝ ﻭﺍﳌﺬﻛﻮﺭ‬
‫ﺃﻋﻼﻩ‪ ،‬ﻭﻣﻨﻪ ﲢﺪﻳﺪ ﺍﻟﺪﺭﺟﺔ ‪ q‬ﻟﻘﺴﻢ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻟﻨﻤﻮﺫﺝ‬
‫)‪.ARMA(p,q‬‬
‫ﺑﻴﻨﻤﺎ ﺇﺫﺍ ﻛﺎﻥ‬
‫‪1‬‬
‫= ) ‪SE ( rk‬‬
‫‪T‬‬
‫ﻓﻨﺴﻤﻴﻪ ﺍﺧﺘﺒﺎﺭ ﺃﻭ ﻗﺎﻋﺪﺓ ﺍﻹ‪‬ﺎﻡ ‪ Rule of thumb‬ﺍﻟﺬﻱ ﻳﺴﺘﻌﻤﻞ ﻣﻊ‪:‬‬
‫ﺍ ‪ -‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻜﺸﻒ ﺇﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪.‬‬

‫‪174‬‬
 ‫ﺍﻟﺸﻜﻞ )‪ :(33‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﻤﺘﻐﲑ ﺍﻻﺟﺘﻤﺎﻋﻲ ﺻﻔﺤﺔ ‪. T=45 ،38‬‬

‫ﺃﻳﻦ ﺍﻋﺘﻤﺪﻧﺎ ﺍﻟﺘﻌﺒﲑ ﻋﻦ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﲟﻨﺤﲏ ﻋﻜﺲ ﺍﳌﺮﺍﺕ ﺍﻟﺴﺎﺑﻘﺔ ﻹﻇﻬﺎﺭ‬
‫ﺍﳋﻄﲔ ﺍﳌﺘﻮﺍﺯﻳﲔ ﺍﳌﻤﺜﻠﲔ ‪‬ﺎﻝ ﺍﻟﺜﻘﺔ ﺍﻟﺬﻱ ﻳﻌﺎﺩﻝ ‪ ، ± 245 » 0. 298‬ﻭ‪‬ﺬﺍ ﻧﺴﺘﻨﺘﺞ‬
‫ﺃﻥ ﻫﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ ﻏﲑ ﻣﺴﺘﻘﺮﺓ ﻛﻮﻥ ﻣﻌﺎﻣﻼﺕ ﺍﻟﺪﺍﻟﺔ ﺃﻋﻼﻩ ﱂ ﺗﺪﺧﻞ ﳎﺎﻝ ﺍﻟﺜﻘﺔ‬
‫ﺑﻌﺪ ﺍﻟﻔﺘﺮﺓ ‪ ،12‬ﻣﻨﻬﺎ ‪. r20 , r18 , r16 , r14‬‬
‫ﺏ ‪ -‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ ﳌﻌﺮﻓﺔ ﺩﺭﺟﺔ ﺍﻟﻘﺴﻢ ﺍﻻﳓﺪﺍﺭﻱ ﻟﻨﻤﻮﺫﺝ‬
‫‪1‬‬
‫= ) ‪SE ( a k‬‬ ‫)‪ ،ARMA(p,q‬ﻭﻫﻮ‬
‫‪T‬‬

‫)‪(A.C‬‬‫‪ -4‬ﻓﺒﺤﺴﺎﺏ ﳎﺎﻝ ﺍﻟﺜﹼﻘﺔ ﻻﺧﺘﺒﺎﺭ ﺑﺎﺭﺗـﻠﺖ ﻭﺗﻄﺒﻴﻘـﻪ ﻋﻠﻰ‬

‫ﻧﺴﺘﻨﺘﺞ ﻣﻌﻨﻮﻳﺔ ‪ r3 , r2 , r1‬ﻓﻘﻂ ﺃﻱ ﺃﻥ ﻗﺴﻢ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻫﻮ‬
‫)‪ MA(3‬ﻭﻣﻨﻪ ‪.q = 3‬‬

‫‪175‬‬
 ‫ﺍﳉﺪﻭﻝ )‪ :(32‬ﻣﻌﻨﻮﻳﺔ ﺍﳌﻌﻠﻤﺎﺕ‬
‫اﻟﻤﻌﻨﻮﯾﺔ‬ ‫) ‪±2SE ( rk‬‬ ‫‪SE ( rk ) 12‬‬ ‫‪rk‬‬ ‫‪k‬‬
‫*‬ ‫‪2‬‬ ‫‪62 =. 25‬‬ ‫‪1‬‬ ‫‪T‬‬ ‫‪0.68‬‬ ‫‪1‬‬
‫*‬ ‫‪0.35‬‬ ‫‪0.18‬‬ ‫‪052‬‬ ‫‪2‬‬
‫*‬ ‫‪0.40‬‬ ‫‪0.20‬‬ ‫‪0.47‬‬ ‫‪3‬‬
‫‪0.43‬‬ ‫‪0.21‬‬ ‫‪0.34‬‬ ‫‪4‬‬
‫‪...‬‬ ‫‪...‬‬ ‫‪...‬‬ ‫‪...‬‬
‫‪0.53‬‬ ‫‪0.26‬‬ ‫‪0.06‬‬ ‫‪16‬‬

‫‪ - 5‬ﺑﻴﻨﻤﺎ ﺣﺴﺎﺏ ﳎﺎﻝ ﺍﻟﺜﹼﻘﺔ ﻟﻼﺧﺘﺒﺎﺭ ﺍﻟﺜﺎﱐ‪ ،‬ﻭﺍﻟﺬﻱ ﳝﺜﻞ ﺍﳋﻄﲔ ﺍﻟﻨﻘﻄﻴﲔ‬
‫ﺍﳌﺘﻮﺍﺯﻳﲔ ﰲ )‪ ،(P.A.C‬ﻧﺴﺘﻨﺘﺞ ﻣﻌﻨﻮﻳﺔ ‪ ، a8 , a1‬ﺃﻱ ﺃﻥ ﺍﻟﻘﺴﻢ ﺍﻻﳓﺪﺍﺭﻱ ﻳﺘﻤﺜﻞ‬
‫ﰲ )‪ ،AR(8‬ﻣﻊ ‪ f 7,... , f 3 , f 2‬ﺗﻌﺘﱪ ﻣﻌﺪﻭﻣﺔ ﻛﻤﺎ ﻳﻌﻜﺴﻪ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪.‬‬

‫‪13‬‬
‫ﻟـ‪.GB13‬‬ ‫ﺍﻟﺸﻜﻞ )‪ :(34‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ‬

‫[ = ) ‪SE ( r3‬‬
‫‪1 + 2{ ( r1 ) 2 + ( r2 ) 2 } 1 2‬‬
‫]‬ ‫‪ - 12‬ﻓﻤﺜ ﹰ‬
‫ﻼ‬
‫‪62‬‬
‫‪ -13‬ﺍﳋﻄﲔ ﺍﳌﺘﻮﺍﺯﻳﲔ ﳝﺜﻼﻥ ﳎﺎﻝ ﺍﻟﺜﻘﺔ ﺍﳌﻌﱪ ﻋﻨﻪ ﺑﺎﺧﺘﺒﺎﺭ ﻗﺎﻋﺪﺓ ﺍﻹ‪‬ﺎﻡ ﺍﳌﺬﻛﻮﺭ ﺳﺎﺑﻘﺎﹰ‪.‬‬

‫‪176‬‬
‫‪ -6‬ﻳﻜﻮﻥ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﺨﺘﺎﺭ ﻋﺒﺎﺭﺓ )‪ ARMA(p,q‬ﺑﻴﻨﻤﺎ ﻳﻜﻮﻥ ﺑﻌﺪ ﺃﺧﺬ‬
‫ﺍﻟﻔﺼﻠﻴﺔ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ ) ‪ SARMA ( p, q ) ( P, Q‬ﺃﻱ ) ‪، SARMA ( 8, 3 )( 1, 1‬‬
‫‪12‬‬ ‫‪s‬‬

‫ﺣﻴﺚ ‪ S‬ﺭﻣﺰ ﻭﺟﻮﺩ ﺍﻟﻔﺼﻠﻴﺔ ﺑﺪﻭﺭﻳﺔ ﻣﻘﺪﺍﺭﻫﺎ )‪ (12‬ﺷﻬﺮﺍﹰ‪ .‬ﳍﺬﺍ ﳚﺐ ﺗﻘﺪﻳﺮ‬

‫ﺍﻟﻨﻤﺎﺫﺝ ﺍﶈﺘﻤﻠﺔ ﻭﺍﳌﺸﺘﻘﺔ ﻣﻦ ﺍﻟﺼﻴﻐﺔ ﺍﻟﻌﺎﻣﺔ ﺃﻋﻼﻩ‪ ،‬ﰒ ﺍﳌﻔﺎﺿﻠﺔ ﺑﻴﻨﻬﺎ ﺑﺎﺳﺘﻌﻤﺎﻝ‬
‫ﺃﺩﻭﺍﺕ ﺇﺧﺘﺒﺎﺭﻳﺔ ﻣﻨﺎﺳﺒﺔ‪.‬‬
‫‪ -7‬ﺑﺎﳋﱪﺓ ﳝﻜﻦ ﻣﺒﺎﺷﺮﺓﹰ ﻭﺑﺎﻟﻨﻈﺮ ﺇﱃ ﺍﻟﺸﻜﻞ )‪ (32‬ﰒ )‪ (34‬ﺍﺳﺘﻨﺘﺎﺝ ﻣﻦ‬
‫ﺍﻷﻭﻝ‪ ،‬ﺃﻥ ﺗﻠﻚ ﺍﻟﺪﺍﻟﺔ ﻗﺪ ﲤﺜﻞ ﺩﺍﻟﺔ ﺍﺭﺗﺒﺎﻁ ﺫﺍﰐ ﻟﻨﻤﻮﺫﺝ )‪ AR(p‬ﺑﺪﻭﺭﻳﺔ ﻣﻘﺪﺍﺭﻫﺎ‬
‫)‪ (12‬ﺷﻬﺮﺍﹰ ﺃﻱ ) ‪ SAR ( p, P‬ﺑﻴﻨﻤﺎ ﺍﻟﺸﻜﻞ ﺍﻟﺜﺎﱐ ﳛﺪﺩ ﺩﺭﺟﺘﻪ ﺃﻱ ‪ p = 1‬ﻟﻴﺼﺒﺢ‬
‫‪12‬‬

‫) ‪. SAR ( 1, 1‬‬
‫‪12‬‬

‫ﺇﻻ ﺃﻧﻨﺎ ﻧﻔﻀﻞ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﺃﻥ ﺗﻨﻌﻜﺲ ﺍﻟﻔﺼﻠﻴﺔ ﰲ ﻗﺴﻢ ﺍﳌﺘﻮﺳﻄﺎﺕ‬

‫ﺍﳌﺘﺤﺮﻛﺔ‪ ،‬ﻛﻮﻥ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ ﺍﻟﺴﺎﻟﻔﺔ ﺍﻟﺬﻛﺮ ﺧﺎﻟﻴﺔ ﻣﻦ ﺩﻟﻴﻞ ﻋﻠﻰ‬
‫ﺍﻟﺘﺄﺛﲑ ﺍﻟﻔﺼﻠﻲ ﻋﻠﻰ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‪.‬‬
‫ﲡﺪﺭ ﺍﻹﺷﺎﺭﺓ ﻫﻨﺎ ﻭﻋﻠﻰ ﻏﺮﺍﺭ ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ‪ ،‬ﺃﻧﻪ ﳝﻜﻦ ﳕﺬﺟﺔ ﺍﻟﻔﺼﻠﻴﺔ‬
‫ﻣﺒﺎﺷﺮﺓ ﻛﻤﺎ ﻭﺭﺩ ﺃﻋﻼﻩ‪ ،‬ﺃﻭ ﺇﺯﺍﻟﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻄﺮﻳﻘﺔ ﻣﻨﺎﺳﺒﺔ ﻛﻤﺎ ﺭﺃﻳﻨﺎ ﻣﻦ ﻗﺒﻞ‪،‬‬
‫ﻓﺘﺤﺪﻳﺪ ﺍﻟﻨﻤﻮﺫﺝ ﰒ ﺇﺭﺟﺎﻉ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﺑﻌﺪ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﺍﳌﻮﻓﻘﺔ‪.‬‬
‫ﺳﻨﺴﺘﻌﺮﺽ ﻧﺘﺎﺋﺞ ﺗﻘﺪﻳﺮ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺍﶈﺪﺩ ﻻﺣﻘﺎﹰ ﺑﻌﺪ ﺗﻨﺎﻭﻝ ﻃﺮﻕ‬
‫ﺍﻟﺘﻘﺪﻳﺮ‪.‬‬
‫‪ - 2.2‬ﻣﺮﺣﻠﺔ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ‬
‫ﺑﻌﺪ ﺍﻻﻧﺘﻬﺎﺀ ﻣﻦ ﻣﺮﺣﻠﺔ ﺍﻟﺘﻌﺮﻑ ﻋﻠﻰ ﳕﻮﺫﺝ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺑﺘﺤﺪﻳﺪ ﻛﻞ‬
‫ﻣﻦ )‪ ،(p, d, q‬ﳝﻜﻦ ﺍﻻﻧﺘﻘﺎﻝ ﺇﱃ ﺍﳌﺮﺣﻠﺔ ﺍﻟﺘﻘﻨﻴﺔ ﺍﳌﻮﺍﻟﻴﺔ ﺍﳌﺘﻤﺜﻠﺔ ﰲ ﺗﻘﺪﻳﺮ‬
‫ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ ‪.‬‬

‫‪177‬‬
 ‫‪ - 1.2.2‬ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﳕﻮﺫﺝ ﺍﳓﺪﺍﺭ ﺫﺍﰐ‬
‫ﰲ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ‪ ،‬ﻭﺑﻌﺪ ﲢﺪﻳﺪ ﺍﻟﺪﺭﺟﺔ ‪ ، p‬ﻳﺼﺒﺢ ﻣﻦ ﺍﳌﻴﺴﻮﺭ‬
‫ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺗﻪ ‪ f p ,....., f 2 , f 1‬ﻭﺫﻟﻚ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺇﺣﺪﻯ ﺍﻟﻄﺮﻕ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫ﺍ ‪ -‬ﻃﺮﻳﻘﺔ ﻣﻌﺎﺩﻻﺕ ﻳﻮﻝ ‪ -‬ﻭﻟﻜﺮ‬
‫ﺗﻠﺠﺄ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺇﱃ ﻣﻌﺎﺩﻻﺕ ﻳﻮﻝ‪-‬ﻭﻟﻜﺮ ﻟﺘﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ‪،‬‬
‫ﺣﻴﺚ ﺃ‪‬ﺎ )ﺍﳌﻘﺪﺭﺍﺕ( ﰲ ﺣﺎﻟﺔ ﳕﺎﺫﺝ ﺍﻝ)‪ AR(p‬ﺗﻜﻮﻥ ﻓﻌﺎﻟﺔ‪ ،‬ﻓﻔﻲ ﺣﺎﻟﺔ )‪AR(2‬‬
‫ﻣﺜﻼﹰ ﺗﻜﻮﻥ ﻟﺪﻳﻨﺎ ﻣﻌﺎﺩﻟﺘﲔ ﻟﻴﻮﻝ ﻭﻭﻟﻜﺮ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪r 1 = f 1 + f 2r1‬‬ ‫)‪(5.33‬‬

‫‪r 2 = f 1r1 + f 2‬‬ ‫)‪(5.34‬‬

‫ﻓﻤﻦ )‪ (5.33‬ﳓﺴﺐ‬
‫) ‪f1 = r1 ( 1 - f 2‬‬ ‫)‪(5.35‬‬

‫ﺑﺎﻟﺘﻌﻮﻳﺾ ﰲ ﺍﳌﻌﺎﺩﻟﺔ ﺍﻟﺜﺎﻧﻴﺔ )‪ (5.34‬ﳓﺴﺐ‬

‫‪r 2 = r 12 ( 1 - f 2 ) + f 2‬‬

‫‪r 2 - r 12‬‬
‫= ‪f2‬‬ ‫)‪(5.36‬‬
‫‪1 - r 12‬‬
‫ﻭﺑﺎﻟﺘﻌﻮﻳﺾ ﰲ )‪ (5.35‬ﻧﺴﺘﻨﺘﺞ‬
‫‪r 2 - r 12‬‬
‫‪f1 = [ 1 -‬‬ ‫‪] r1‬‬
‫‪1 - r 12‬‬

‫‪178‬‬
‫ﻭﺑﺘﻌﻮﻳﺾ ﻣﻌﻠﻤﱵ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺑﺎﳌﻌﻠﻤﺘﲔ ﺍﳋﺎﺻﺔ ﺑﺎﻟﻌﻴﻨﺔ ﳓﺼﻞ‬
‫ﻋﻠﻰ‪:‬‬
‫^‬ ‫‪r2 - r12‬‬
‫‪f1 = [ 1 -‬‬ ‫‪] r1‬‬ ‫)‪(5.37‬‬
‫‪1 - r12‬‬

‫‪1 - r2‬‬
‫[=‬ ‫‪] r1‬‬
‫‪1 - r12‬‬

‫^‬ ‫‪r2 - r12‬‬

‫= ‪f2‬‬ ‫)‪(5.38‬‬
‫‪1 - r12‬‬

‫ﻭﻳﺘﻢ ﺍﳊﻞ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺼﻔﻮﻓﺎﺕ ﰲ ﺣﺎﻟﺔ )‪ AR(3‬ﻣﺜﻼﹰ ﻛﻤﺎ ﻳﻠﻲ ‪:‬‬

‫‪é ^ ù‬‬
‫‪é r1 ù é 1‬‬ ‫‪r1‬‬ ‫‪r2 ù ê f 1 ú‬‬
‫‪êr ú = ê r‬‬ ‫‪1‬‬ ‫‪r1 úú * êf 2 ú‬‬
‫^‬

‫‪ê 2ú ê 1‬‬ ‫‪ê ^ ú‬‬

‫‪êë r3 úû êë r2‬‬ ‫‪r1‬‬ ‫‪1 úû êf ú‬‬
‫‪ëê ûú‬‬
‫‪3‬‬

‫ﺃﻱ ﻭﺑﺸﻜﻞ ﳐﺘﺼﺮ‬

‫^‬
‫‪R=A‬‬ ‫‪´F‬‬
‫^‬
‫‪F = A -1 R‬‬ ‫ﻭﻣﻨﻪ‬

‫ﺏ ‪ -‬ﺍﻟﻄﺮﻳﻘﺔ ﺍﻻﳓﺪﺍﺭﻳﺔ‬
‫ﻟﺘﻮﺿﻴﺢ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻧﻔﺘﺮﺽ ﳕﻮﺫﺝ )‪ ،AR(2‬ﻭﺑﺴﺒﺐ ﻣﺸﻜﻞ ﻗﻴ‪‬ﻢ‬
‫ﺍﻻﻧﻄﻼﻕ‪ ،‬ﻧﺒﺪﺃ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻘﺪﻳﺮ ﻣﻦ ﺍﻟﻔﺘﺮﺓ ) ‪.( t = p + 1 = 3‬‬
‫‪y t = d + f 1 y t -1 + f 2 y t - 2 + e t‬‬

‫‪179‬‬
 ‫ﻭﺑﺎﻟﺘﻌﻮﻳﺾ‬
‫‪y 3 = d + f1 y 2 + f 2 y 1 + e 3‬‬
‫‪y 4 = d + f1 y 3 + f 2 y 2 + e4‬‬
‫‪y T = d + f 1 y T- 1 + f 2 y T- 2 + e T‬‬

‫ﻭﺑﻜﺘﺎﺑﺘﻬﺎ ﰲ ﺷﻜﻞ ﻣﺼﻔﻮﻓﺎﺕ ﻓﺈ‪‬ﺎ ﺗﺄﺧﺬ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ ﺑﺎﻷﺑﻌﺎﺩ ﺍﳌﺪﻭﻧﺔ‬

‫ﲢﺖ ﺍﻟﻜﺘﺎﺑﺔ ﺍﳌﺨﺘﺼﺮﺓ ﺍﳌﺮﻓﻘﺔ‪:‬‬

‫‪é y3 ù‬‬ ‫‪é1‬‬ ‫‪y2‬‬ ‫‪y1 ù‬‬ ‫‪ée 3 ù‬‬

‫‪êy ú‬‬ ‫‪ê1‬‬ ‫‪ú‬‬ ‫‪éd ù ê ú‬‬
‫= ‪ê 4ú‬‬ ‫‪ê‬‬ ‫‪y3‬‬ ‫‪y 2 ú ê ú êe 4 ú‬‬
‫‪* f +‬‬
‫‪ê .. ú‬‬ ‫‪ê..‬‬ ‫‪..‬‬ ‫‪.. ú ê 1 ú ê .. ú‬‬
‫‪ê ú‬‬ ‫‪ê‬‬ ‫‪ú êf ú ê ú‬‬
‫‪ë yT û‬‬ ‫‪ë1‬‬ ‫‪y T -1‬‬ ‫‪y T - 2 û ë 2 û ëe T û‬‬
‫‪Y‬‬ ‫=‬ ‫‪X‬‬ ‫*‬ ‫‪F‬‬ ‫‪+‬‬ ‫‪E‬‬
‫]‪[(T-P).1‬‬ ‫] ‪[(T-P).3‬‬ ‫] ‪[3 . 1‬‬ ‫]‪[(T-P).1‬‬

‫ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻭﲢﺖ ﻓﺮﺿﻴﺎﺕ ﻣﻌﺮﻭﻓﺔ‪ ،‬ﳝﻜﻦ ﺗﻘﺪﻳﺮ ﺷﻌﺎﻉ ﺍﳌﻘﺪﺭﺍﺕ‬

‫ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﺍﻟﻌﺎﺩﻳﺔ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫ﺃﻳﻦ‬
‫‪éd ù‬‬
‫‪F = êêf1 úú‬‬
‫‪êëf 2 úû‬‬

‫^‬
‫‪F = ( X ' X ) -1 X ' Y‬‬ ‫)‪(5.39‬‬

‫‪180‬‬
 ‫‪Maximum Likelihood‬‬ ‫ﺟـ‪ -‬ﻃﺮﻳﻘﺔ ﺍﳌﻌﻘﻮﻟﻴﺔ ﺍﻟﻌﻈﻤﻰ‪:‬‬
‫ﻳﺘﻮﻗﻒ ﺍﻟﺘﻘﺪﻳﺮ ‪‬ﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺃﺳﺎﺳﺎﹰ ﻋﻠﻰ ﲢﻘﻖ ﻓﺮﺿﻴﺔ ﺍﻟﺘﻮﺯﻳﻊ ﺍﻟﻄﺒﻴﻌﻲ‪.‬‬
‫ﺍﻟﱵ ﺗﻌﺘﻤﺪ ﻣﺒﺪﺃ ﺗﺼﻐﲑ ﺃﻭ ﺗﺪﻧﻴﺔ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ‪ ،Min RSS‬ﲟﻌﲎ‬
‫ﺍﺧﺘﻴﺎﺭ ﺷﻌﺎﻉ ﺍﳌﻌﻠﻤﺎﺕ ) ‪ ( f p ,....., f 2 , f 1‬ﺍﻟﺬﻱ ﻳﻀﻤﻦ ﺗﺼﻐﲑ ﳎﻤﻮﻉ‬
‫ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‪ ،‬ﺃﻱ‪:‬‬
‫‪Min S ( f$ ) = å e 2t‬‬

‫ﳝﻜﻦ ﺍﻻﺳﺘﻌﺎﻧﺔ ‪‬ﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻋﻨﺪ ﺗﻘﺪﻳﺮ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ‪ .‬ﻳﺘﻢ ﰲ ﺗﻠﻚ‬
‫ﺍﳊﺎﻟﺔ ﺍﺧﺘﻴﺎﺭ ﻣﻘﺪﺭﺍﺕ ﺷﻌﺎﻋﻲ ﺍﳌﻌﻠﻤﺎﺕ ﺍﳋﺎﺻﺔ ﺑﺎﻟﻘﺴﻤﲔ ﺍﻻﳓﺪﺍﺭﻱ ﺃﻭ‬
‫ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ) ‪ F = ( f 1 , f 2 ,.... f p‬ﻭ ) ‪ Q = ( q 1 , q 2 ,.... q p‬ﻋﻠﻰ‬
‫ﺍﻟﺘﺮﺗﻴﺐ‪ ،‬ﺍﻟﱵ ﺗﻀﻤﻦ ﺗﺼﻐﲑ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ﻛﺎﻟﻌﺎﺩﺓ‬

‫‪Min S ( f$ , q$ ) = å e2t‬‬

‫‪$ -1 ( L ) F‬‬
‫‪et = Q‬‬ ‫‪$ ( L )y‬‬
‫‪t‬‬ ‫ﺣﻴﺚ‬

‫‪yt‬‬ ‫ﻧﺸﲑ ﺇﱃ ﺃﻥ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﲢﺘﺎﺝ ﺇﱃ ﺗﻮﻓﲑ ﻗﻴ‪‬ﻢ ﺍﺑﺘﺪﺍﺋﻴﺔ ﺧﺎﺻﺔ ﺑﺎﳌﺘﻐﲑ‬
‫ﻣﺜﻞ ‪ y 0‬ﻭ ‪ y -1‬ﻭ‪ y - p ...‬ﺣﻴﺚ ﺩﺍﻟﺔ ﺍﳌﻌﻘﻮﻟﻴﺔ ﺍﻟﻌﻈﻤﻰ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﺗﻜﻮﻥ‬
‫ﺷﺮﻃﻴﺔ ﳍﺬﺍ ﺍﻟﺴﺒﺐ‪ .‬ﻭﳝﻜﻦ ﻓﻬﻢ ﻫﺬﻩ ﺍﻟﻈﺎﻫﺮﺓ ﺑﺴﻬﻮﻟﺔ ﻋﻨﺪ ﺗﻌﻮﻳﺾ ‪t‬‬
‫ﺑـ ) ‪ ( p,..., 2, 1‬ﰲ ﺩﺍﻟﺔ ﺍﳌﻌﻘﻮﻟﻴﺔ ﺍﻟﻌﻈﻤﻰ ﺃﻭ ﰲ ﻋﻼﻗﺔ ﺍﻟﺒﻮﺍﻗﻲ ﺃﻋﻼﻩ‪.‬‬

‫‪ -2.2.2‬ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﻭﺍﳌﺨﺘﻠﻄﺔ ‪:‬‬

‫ﺗﻌﺘﱪ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ )‪ MA(q‬ﻭ )‪ ARMA(p,q‬ﺃﻋﻘﺪ ﺑﻜﺜﲑ ﻣﻦ ﺣﻴﺚ‬
‫ﺍﻟﺘﻘﺪﻳﺮ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻻﳓﺪﺍﺭﻳﺔ‪ ،‬ﻛﻮ‪‬ﺎ ﻏﲑ ﺧﻄﻴﺔ ﰲ ﺍﳌﻌﻠﻤﺎﺕ ﻣﻦ ﺟﻬﺔ ﻭﻋﺪﻡ‬
‫ﻣﺸﺎﻫﺪﺓ ﻣﺘﻐﲑ ﺍﻷﺧﻄﺎﺀ ﻣﻦ ﻧﺎﺣﻴﺔ ﺛﺎﻧﻴﺔ‪.‬‬

‫‪181‬‬
‫ﻓﻬﺪﻑ ﺍﻟﺘﻘﺪﻳﺮ ﻳﺘﻜﺜﻞ ﰲ ﲢﺪﻳﺪ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻘﺴﻢ ﺍﻻﳓﺪﺍﺭﻱ ﻭﻗﺴﻢ‬
‫ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ )‪ ARMA(p,q‬ﻣﻌﺎﹰ‪ ،‬ﺃﻭ ﻣﻌﻠﻤﺎﺕ ﻗﺴﻢ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‬
‫ﻟﻮﺣﺪﻫﺎ ﰲ ﳕﻮﺫﺝ )‪.MA(q‬‬
‫ﻓﻔﻲ ﺣﺎﻟﺔ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﺨﺘﻠﻂ ﺍﻟﻌﺎﻡ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪y t - f 1 y t - 1 - ... - f p y t - p = e t + q 1 e t - 1 +... + q q e t - q‬‬

‫‪F( L ) y t = Q( L )e t‬‬ ‫ﺃﻱ‬

‫ﺣﻴﺚ‬
‫‪F( L ) = 1 - f 1 L - f 2 L 2 -... - f p L p‬‬
‫‪Q(L) = 1+ q1L + q2 L2 + ...+ qq Lq‬‬

‫ﺑﺎﻓﺘﺮﺍﺽ ﺇﻣﻜﺎﻧﻴﺔ ﻗﻠﺐ ﺍﳌﻌﺎﻣﻞ ) ‪ Q( L‬ﻓﺈﻥ‬

‫‪e t = Q -1 ( L ) F ( L ) y t‬‬

‫ﳍﺬﺍ ﻓﺈﻥ ﺃﻱ ﻃﺮﻳﻘﺔ ﺗﻘﺪﻳﺮ‪ ،‬ﳚﺐ ﺃﻥ ﺗﺄﺧﺬ ﺑﻌﲔ ﺍﻻﻋﺘﺒﺎﺭ ﻓﻜﺮﺓ ﺗﺪﻧﻴﺔ‬
‫ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‪ ،‬ﺃﻱ‬
‫) ‪å e 2t = s ( f , q‬‬
‫‪t‬‬

‫ﻭﺑﺎﻟﺘﺎﱄ ﻓﺈﻥ‬
‫^‬ ‫^‬
‫) ‪å e 2t = s ( f , q‬‬
‫‪t‬‬

‫ﺣﻴﺚ‬
‫^‬ ‫^‬
‫‪-1‬‬
‫‪e t = Q ( L ) F( L ) y t‬‬ ‫)‪(5.40‬‬

‫ﻟﻘﺪ ﺭﺃﻳﻨﺎ ﺇﻣﻜﺎﻧﻴﺔ ﻭﺳﻬﻮﻟﺔ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﻫﺬﻩ ﺍﻟﻌﻼﻗﺔ ﰲ ﺣﺎﻟﺔ ﻏﻴﺎﺏ‬
‫ﺍﻟﻄﺮﻑ )‪ ،MA(q‬ﺑﻴﻨﻤﺎ ﺗﺼﺒﺢ ﰲ ﺣﺎﻟﺔ ﺣﻀﻮﺭﻩ ﻟﻮﺣﺪﻩ ﺃﻭ ﻣﻊ ﻣﺮﻛﺒﺔ ﺍﻟﻨﻤﺎﺫﺝ‬

‫‪182‬‬
‫ﺍﻻﳓﺪﺍﺭﻳﺔ )‪ ،AR(p‬ﻏﲑ ﺧﻄﻴﺔ ﺍﳌﻌﻠﻤﺎﺕ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﺗﺘﻄﻠﺐ ﻃﺮﻳﻘﺔ ﺗﻘﺪﻳﺮ ﺗﻜﺮﺍﺭﻳﺔ‬
‫)‪ .(Non linear Iterative Routine‬ﻭﻣﻦ ﺑﲔ ﻫﺬﻩ ﺍﻟﻄﺮﻕ‪:‬‬

‫‪ - 1 - 2‬ﻃﺮﻳﻘﺔ ﺍﻟﺒﺤﺚ ﺍﻟﺸﺒﻜﻲ ‪: Grid - Search‬‬

‫ﻟﺘﻮﺿﻴﺤﻬﺎ ﻧﺪﺭﺝ ﺍﳌﺜﺎﻝ ﺍﳌﺨﺘﻠﻂ ﺍﻟﺒﺴﻴﻂ ﺍﻟﺘﺎﱄ ) ‪ARMA( 1,1‬‬

‫‪y t - fy t - 1 = e t + qe t - 1‬‬

‫‪( 1 - fL ) y t = e t + q e t - 1‬‬ ‫ﺇﺫﺍﹰ‬

‫‪y‬‬ ‫‪t‬‬ ‫=‬

‫‪1‬‬
‫) ‪( e t + q e t-1‬‬ ‫)‪(5.41‬‬ ‫ﻭﻣﻨﻪ‬
‫) ‪( 1 - fL‬‬

‫ﺑـ ‪vt‬‬ ‫‪1‬‬

‫‪et‬‬ ‫ﻓﺈﺫﺍ ﲰﻴﻨﺎ‬
‫) ‪( 1 - fL‬‬

‫= ‪vt‬‬
‫‪1‬‬
‫‪et‬‬ ‫ﺃﻱ‬
‫) ‪( 1 - fL‬‬

‫ﻭﺑﻌﻤﻠﻴﺔ ﺑﺴﻴﻄﺔ ﻓﺈﻥ‬

‫‪v t = fv t - 1 + e t‬‬ ‫)‪(5.42‬‬

‫ﻧﻼﺣﻆ ﻋﻨﺪ ﺍﻟﻌﻼﻗﺔ ﺍﻷﺧﲑﺓ ﻫﺬﻩ‪ ،‬ﺃﻧﻪ ﻟﻮ ﺗﻮﻓﺮﺕ ﻗﻴ‪‬ﻢ ﺍﻟﺸﻌﺎﻉ ‪ ، vt‬ﳝﻜﻦ‬
‫ﺗﻘﺪﻳﺮ ﺍﳌﻌﻠﻤﺔ ‪ f‬ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ‪ .‬ﻟﻜﻦ ﺑﺴﺒﺐ ﻋﺪﻡ ﻣﺸﺎﻫﺪ‪‬ﺎ‪ ،‬ﻧﻠﺠﺄ‬
‫ﺇﱃ ﺍﻟﻌﻤﻠﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ ﺣﻴﺚ ﻧﺮﺟﻊ ﺇﱃ ﺍﻟﻌﻼﻗﺔ )‪ (5.41‬ﻭﻧﻌﻴﺪ ﻛﺘﺎﺑﺘﻬﺎ ﰲ ﺍﻟﺸﻜﻞ‬
‫ﺍﳌﻮﺳﻊ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪1‬‬ ‫‪q‬‬
‫= ‪yt‬‬ ‫‪et +‬‬ ‫‪e t -1‬‬
‫) ‪( 1 - fL‬‬ ‫) ‪( 1 - fL‬‬

‫‪y t = v t + qv t -1‬‬ ‫ﻭﻣﻨﻪ‬

‫‪183‬‬
‫ﻣﻦ ﻫﺬﻩ ﺍﳌﻌﺎﺩﻟﺔ ﻭﺑﺘﻌﻮﻳﺾ ‪ q‬ﺑﻘﻴ‪‬ﻤﻬﺎ ﺍﻟﱵ ﺗﻘﻊ ﺿﻤﻦ ﺍ‪‬ﺎﻝ ‪ q < 1‬ﻣﻦ‬
‫ﺃﺟﻞ ﺷﺮﻁ ﺃﻣﻜﺎﻧﻴﺔ ﻗﻠﺐ ﺍﻟﻨﻤﻮﺫﺝ " ‪ ،"Invertibility‬ﻭﺑﺘﻮﻓﲑ ﺍﻟﻘﻴ‪‬ﻢ ﺍﻻﺑﺘﺪﺍﺋﻴﺔ‬
‫ﻝ ‪ vt‬ﺃﻭ ﺟﻌﻠﻬﺎ ﻣﺴﺎﻭﻳﺔ ﻟﻠﺼﻔﺮ‪ ،‬ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﺃﻭ ﺍﳌﺜﺎﻝ ‪ ، v 0 = 0‬ﳓﺼﻞ ﻋﻠﻰ‪:‬‬

‫‪v t = y t - qv t - 1‬‬

‫ﺇﺫﺍﹰ ﺗﺒﺪﺃ ﺍﻟﻌﻤﻠﻴﺔ ﺑﺎﻟﺘﻜﺮﺍﺭ ﺍﻷﻭﻝ ﺫﻟﻚ ﺑﺎﺧﺘﻴﺎﺭ ﻣﺜﻼ ‪ q = -0. 9‬ﻭﻧﺴﻤﻴﻬﺎ‬
‫‪14‬‬

‫) ‪ q( 1‬ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪t=1‬‬ ‫‪:‬‬ ‫‪v (11 ) = y 1‬‬

‫‪t=2‬‬ ‫‪:‬‬ ‫) ‪v (21 ) = y 2 - q ( 1 ) v (11‬‬

‫) ‪t = 3 : v (31 ) = y 3 - q ( 1 ) v (21‬‬

‫‪........‬‬
‫‪t=T‬‬ ‫‪:‬‬ ‫‪v(T1 ) = y T - q ( 1 ) v(T1-)1‬‬

‫‪év1(1) ù‬‬
‫‪ê (1) ú‬‬
‫‪v‬‬
‫)‪V (1‬‬ ‫‪=ê 2 ú‬‬ ‫ﺣﻴﺚ‬
‫‪ê ... ú‬‬
‫‪ê (1) ú‬‬
‫‪ëêvT ûú‬‬

‫ﺑﺘﻌﻮﻳﺾ ﻫﺬﺍ ﺍﻟﺸﻌﺎﻉ ﺍﻟﻨﺎﺗﺞ ﰲ ﺍﳌﻌﺎﺩﻟﺔ )‪ (43‬ﳝﻜﻦ ﺗﻘﺪﻳﺮ ﺍﳌﻌﻠﻤﺔ ‪ f‬ﺃﻱ‬

‫‪ - 14‬ﻣﺎ ﺑﲔ ﻗﻮﺳﲔ ﳝﺜﻞ ﺩﻟﻴﻞ ﺍﻟﺘﻜﺮﺍﺭ‪ ،‬ﰲ ﻫﺬﻩ ﺍﳌﺮﺓ ﻣﻌﻨﺎﻩ ﺍﻟﺘﻜﺮﺍﺭ ﺍﻷﻭﻝ‪.‬‬

‫‪184‬‬
 ‫)‪(1) (1‬‬
‫)‪^ (1‬‬ ‫‪å v t v t-1‬‬
‫‪f‬‬ ‫=‬ ‫‪t‬‬ ‫)‪(5.43‬‬
‫‪å ( v (t 1- 1) ) 2‬‬
‫‪t‬‬

‫ﺫﻟﻚ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻄﺮﻳﻘﺔ ﺍﳋﻄﻴﺔ ﻟﻠﺘﻘﺪﻳﺮ )‪.(O.L.S‬‬

‫ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﺣﺴﺎﺏ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ﺍﳌﻘﺎﺑﻠﺔ ﻟﻠﻤﻌﻠﻤﺘﲔ‬
‫)‪^ (1‬‬
‫) ‪ ( q ( 1 ) , f‬ﻭﻣﻦ ﺍﻟﻌﻼﻗﺔ )‪ (5.42‬ﻛﺎﻵﰐ‪:‬‬
‫)‪^ (1‬‬
‫‪å et2‬‬ ‫‪= å (vt(1) - f vt(1-)1 )2‬‬
‫)‪(1‬‬

‫‪t‬‬ ‫‪t‬‬

‫ﻧﺴﻤﻲ ﳎﻤﻮﻉ ﺍﳌﺮﺑﻌﺎﺕ ﻫﺬﻩ ﺑﺎﻟﺮﻣﺰ ﺍﳌﺘﻌﺎﺭﻑ ﻋﻠﻴﻪ ﺍﳌﻮﺍﻓﻖ ﻟﻠﺘﻜﺮﺍﺭ ﺍﻷﻭﻝ‬
‫) ‪. RSS ( 1‬‬
‫ﻧﻌﻴﺪ ﺍﻟﻌﻤﻠﻴﺔ ﻟﻠﻤﺮﺓ ﺍﻟﺜﺎﻧﻴﺔ )ﺍﻟﺘﻜﺮﺍﺭ ﺍﻟﺜﺎﱐ( ﺣﻴﺚ ﻭﻭﻓﻖ ﺍﳌﺮﺍﺣﻞ ﺍﻟﺴﺎﺑﻘﺔ‬
‫ﻭﺍﻟﱵ ﳔﺘﺼﺮﻫﺎ ﻓﻴﻤﺎ ﻳﻠﻲ‪:‬‬
‫ﻣﺜﻼﹰ‬ ‫ﺍﳌﻮﺍﻟﻴﺔ ‪q = -. 08‬‬ ‫‪-‬ﺣﺴﺎﺏ ﺍﻟﺸﻌﺎﻉ ) ‪ V ( 2‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﻗﻴﻤﺔ‬
‫ﻭﺑﺎﻋﺘﻤﺎﺩ ﺧﻄﻮﺓ )‪ (Increment‬ﻣﻘﺪﺍﺭﻫﺎ )‪.(0.1‬‬
‫‪ -‬ﺗﻘﺪﻳﺮ ﺍﳌﻌﻠﻤﺔ‬
‫)‪( 2) ( 2‬‬
‫)‪^ ( 2‬‬ ‫‪å v t v t -1‬‬
‫‪f‬‬ ‫=‬ ‫‪t‬‬
‫‪( 2) 2‬‬
‫) ‪å ( v t -1‬‬
‫‪t‬‬

‫‪ -‬ﺣﺴﺎﺏ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ﻭﻛﻤﺎ ﰲ ﺍﻟﺴﺎﺑﻖ‬

‫)‪^ (1‬‬
‫‪RSS ( 2 ) = å ( v(t 2 ) - f‬‬ ‫‪v(t 2- 1) ) 2‬‬
‫‪t‬‬

‫ﻧﻜﺮﺭ ﺍﻟﻌﻤﻠﻴﺔ ﻫﺬﻩ ﺣﱴ ﻧﻐﻄﻲ ﻛﺎﻣﻞ ﳎﺎﻝ ﺍﻟﺘﻌﻮﻳﺾ ﻝ‪ ،q‬ﻭﺣﱴ ﳓﺼﻞ‬

‫ﻋﻠﻰ ﺍﳌﻌﻠﻤﺘﲔ ﰲ ﻫﺬﺍ ﺍﳌﺜﺎﻝ ﺍﻟﱵ ﺗﺪﱐ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‪ .‬ﰲ ﻣﺜﺎﻟﻨﺎ ﺍﻟﺒﺴﻴﻂ‬
‫ﻫﺬﺍ‪ ،‬ﻳﻜﻮﻥ ﻋﺪﺩ ﺍﻟﺘﻜﺮﺍﺭﺍﺕ ﻳﻌﺎﺩﻝ ‪ 18‬ﻣﺮﺓ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ‪:‬‬

‫‪185‬‬
 ‫ﺍﳉﺪﻭﻝ)‪ :(33‬ﺗﻘﺪﻳﺮ ﺍﳌﻌﻠﻤﺎﺕ‬
‫ﻡ‪.‬ﻡ‪.‬ﺍﻟﺒﻭﺍﻗﻲ‬ ‫‪f‬‬ ‫‪q‬‬ ‫ﺍﻟﺘﻜﺭﺍﺭ‬
‫)‪^ ( 1‬‬
‫) ‪RSS ( 1‬‬ ‫‪f‬‬ ‫‪0.9-‬‬ ‫‪1‬‬
‫)‪^ ( 2‬‬
‫) ‪RSS ( 2‬‬ ‫‪f‬‬ ‫‪0.8-‬‬ ‫‪2‬‬
‫‪...‬‬ ‫‪...‬‬ ‫‪...‬‬ ‫‪...‬‬
‫) ‪^ ( 18‬‬
‫) ‪RSS ( 18‬‬ ‫‪f‬‬ ‫‪0.9+‬‬ ‫‪18‬‬

‫)‪^(i‬‬
‫ﺍﻟﱵ ﺗﻘﺎﺑﻠﻬﺎ ﺃﺻﻐﺮ‬ ‫‪(q‬‬ ‫)‪(i‬‬
‫‪,f‬‬ ‫)‬ ‫ﻣﻦ ﻫﺬﺍ ﺍﳉﺪﻭﻝ ﳔﺘﺎﺭ ﻛﻤﻘﺪﺭﺍﺕ‬
‫) ‪ RSS ( i‬ﻭﺍﳌﻮﺍﻓﻘﺔ ﻟﻠﺘﻜﺮﺍﺭ )‪.(i‬‬
‫ﻧﺸﲑ ﺇﱃ ﺃﻥ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺗﺼﺒﺢ ﻏﲑ ﻣﺮﻏﻮﺑﺎﹰ ﻓﻴﻬﺎ ﳌﺎ ﻳﺘﺠﺎﻭﺯ ﻋﺪﺩ‬
‫ﻣﻌﻠﻤﺎﺕ ﻗﺴﻢ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ ﺍﻻﺛﻨﲔ )‪ ( 2‬ﺃﻱ ‪ q > 2‬ﻭﺫﻟﻚ ﻟﺼﻌﻮﺑﺔ‬
‫ﻋﻤﻠﻴﺔ ﺍﳊﺴﺎﺏ ﻣﻦ ﺟﻬﺔ ﻭﻛﺬﺍ ﻋﺪﻡ ﺍﺗﺴﺎﻕ ﺍﳌﻌﻠﻤﺎﺕ ﻣﻦ ﺟﻬﺔ ﺛﺎﻧﻴﺔ‪.‬‬

‫‪Gauss-Newton‬‬ ‫‪ -2 -2‬ﻃﺮﻳﻘﺔ ﻏﻮﺱ_ﻧﻴﻮﺗﻦ ﺍﻟﺘﻜﺮﺍﺭﻳﺔ‪:‬‬

‫ﺗﻌﺘﻤﺪ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﺃﻳﻀﺎﹰ ﻋﻠﻰ ﺗﺪﻧﻴﺔ ﳎﻤﻮﻉ ﻣﺮﺑ‪‬ﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‬

‫] ‪Min S ( f, q ) = å e2t = å [ e t / f, q, y t‬‬

‫‪t‬‬ ‫‪t‬‬

‫‪et = Q‬‬ ‫‪-1‬‬

‫‪( L) F (L) y t‬‬ ‫ﺣﻴﺚ‬

‫ﻓﻔﻲ ﺣﺎﻟﺔ ﺍﻟـ)‪ ARMA(1,1‬ﺗﺘﺒﺴﻂ ﻫﺬﻩ ﺍﻟﻌﻼﻗﺔ ﺇﱃ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪1-fL‬‬
‫= ‪et‬‬ ‫‪yt‬‬
‫‪1+qL‬‬

‫‪186‬‬
 ‫^‬ ‫^‬ ‫^‬ ‫^‬
‫ﻟﻠﺤﺼﻮﻝ ﻋﻠﻰ ‪ f‬ﻭ ‪ q‬ﺍﻟﱵ ﺗﺪﻧ‪‬ﻲ ﺍﻟﺪﺍﻟﺔ ) ‪ S ( f , q‬ﳚﺐ ﺍﻟﻠﱡﺠﻮﺀ ﺇﱃ‬
‫ﻣﻨﺸﻮﺭ ﺗﻴﻠﺮ )‪ (Taylor‬ﺣﻮﻝ ﻗﻴﻤﺔ ﺍﺑﺘﺪﺍﺋﻴﺔ ﻣﻌﻴﻨﺔ‪ .‬ﻓﻔﻲ ﺣﺎﻟﺔ ﺍﻟﺪﺍﻟﺔ ) ‪ f ( x‬ﻭﺍﻟﱵ‬
‫ﻧﺮﻳﺪ ﻧﺸﺮﻫﺎ ﺣﻮﻝ ﻗﻴﻤﺔ ﺛﺎﺑﺘﺔ ‪ ، x 0‬ﻓﺈﻥ ﻫﺬﺍ ﺍﳌﻨﺸﻮﺭ ﺍﻟﺜﻨﺎﺋﻲ ﻳﻜﻮﻥ ﻋﺒﺎﺭﺓ ﻋﻦ‪:‬‬

‫‪¶f ( x‬‬ ‫)‬ ‫) ‪1 ¶f ( x 0‬‬

‫‪f( x ) = f( x‬‬ ‫‪0‬‬ ‫‪)+‬‬ ‫‪0‬‬
‫‪( x - x‬‬ ‫‪0‬‬ ‫‪)+‬‬ ‫‪( x - x‬‬ ‫‪0‬‬ ‫‪) 2 + rn‬‬
‫‪¶x‬‬ ‫‪2 ! ¶x 2‬‬

‫)‪.(Reminder‬‬ ‫ﺣﻴﺚ ‪ rn‬ﳝﺜﻞ ﺍﳌﻘﺪﺍﺭ ﺍﳌﺘﺒﻘﻲ‬

‫ﰲ ﻣﻨﺸﻮﺭ‬ ‫‪b0‬‬ ‫ﺣﻮﻝ ﺍﻟﻘﻴﻤﺔ‬ ‫) ‪f( x t ,b‬‬ ‫ﺃﻣﺎ ﰲ ﺣﺎﻟﺔ ﺍﻷﺷﻌﺔ‪ ،‬ﻓﻨﺸﺮ ﺍﻟﺪﺍﻟﺔ‬
‫ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﻳﺄﺧﺬ ﺍﻟﺸﻜﻞ ‪:‬‬
‫‪15‬‬

‫) ‪¶f ( xt , b‬‬
‫‪b0‬‬
‫‪f ( xt , b ) = f ( xt , b 0 ) +‬‬ ‫‪(b - b 0 ) + rn‬‬
‫‪¶b‬‬
‫‪¶ f‬‬ ‫‪é¶ f‬‬ ‫‪¶ f ù‬‬
‫‪= Db = ê‬‬ ‫‪. .‬‬ ‫‪ú‬‬ ‫ﺣﻴﺚ‬
‫‪¶ b‬‬ ‫‪ë ¶ b1‬‬ ‫‪¶ bk û‬‬

‫ﳝﺜﻞ ﻋﺪﺩ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ‪.‬‬ ‫‪k‬‬ ‫ﺑﻴﻨﻤﺎ‬

‫ﰲ ﺣﺎﻟﺔ ﺩﺍﻟﺔ ﻏﲑ ﺧﻄﻴﺔ ﻣﻦ ﺍﻟﺸﻜﻞ‬
‫‪y t = f( x t,b ) + u t‬‬

‫ﺑﺘﻌﻮﻳﺾ ﺍﳌﻨﺸﻮﺭ ﺍﻟﺴﺎﺑﻖ ﰲ ﺍﻟﺪﺍﻟﺔ ﺃﻋﻼﻩ ﳓﺼﻞ ﻋﻠﻰ‬

‫) ‪¶f ( x t , b‬‬
‫‪b0‬‬
‫‪y t = f ( x t , b0 ) +‬‬ ‫‪( b - b 0 ) + rn + u t‬‬
‫‪¶b‬‬

‫‪.b 0‬‬ ‫‪ ¶f ( xt ,b)b‬ﻧﻌﲏ ‪‬ﺬﺍ ﺍﳌﻘﺪﺍﺭ ﺃﻥ ﺍﳌﺸﺘﻘﺔ ﺍﻷﻭﱃ ﻣﻘﻮ‪‬ﻣﺔ ) ‪ (Evaluated at b‬ﻋﻨﺪ ﺍﻟﻨﻘﻄﺔ‬
‫‪0‬‬
‫‪-15‬‬
‫‪0‬‬

‫‪187‬‬
‫ﻭﺑﺈﳘﺎﻝ ﺍﳌﻘﺪﺍﺭ ﺍﳌﺘﺒﻘﻲ ﻣﻦ ﺍﳌﻌﺎﺩﻟﺔ ﻭﻛﺘﺎﺑﺘﻬﺎ ﰲ ﺷﻜﻞ ﺧﻄﻲ ﺑﻄﺮﻳﻘﺔ‬
‫ﺍﳌﺼﻔﻮﻓﺎﺕ‪:‬‬
‫‪Y = F( x , b0 ) + G( x , b0 )( b - b0 ) + U‬‬

‫‪é y1 ù‬‬ ‫‪é u1 ù‬‬

‫‪êy ú‬‬ ‫‪êu ú‬‬
‫‪Y = ê 2ú‬‬ ‫‪,‬‬ ‫‪U = ê 2ú‬‬ ‫ﺣﻴﺚ‬
‫‪ê . ú‬‬ ‫‪ê.ú‬‬
‫‪ê ú‬‬ ‫‪ê ú‬‬
‫‪ë yT û‬‬ ‫‪ëuT û‬‬
‫) ‪G( x , b ) = g ( x 1 , b ) g ( x 2 , b ) ... g ( x T , b‬‬

‫ﻫﻮ ) ) ‪. ( T ´ ( p + q‬‬ ‫ﻝ ) ‪G( x , b‬‬ ‫ﺃﻳﻦ ﻳﻜﻮﻥ ﺍﻟﺒﻌﺪ ﺍﳌﺼﻔﻮﻓﺎﰐ‬

‫ﻟﻐﺮﺽ ﺍﻟﺘﻘﺪﻳﺮ ﺍﳋﻄﻲ ﳍﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ‪ ،‬ﳝﻜﻦ ﺇﻋﺎﺩﺓ ﺗﺮﺗﻴﺐ ﺍﻟﻌﻼﻗﺔ ﺍﳋﻄﻴﺔ‬
‫ﺃﻋﻼﻩ ﰲ ﺍﻟﺸﻜﻞ‪:‬‬
‫‪Y - F( x , b 0 ) = G ( x , b 0 )( b - b 0 ) + U‬‬

‫ﺍﻟﱵ ﳝﻜﻦ ﺻﻴﺎﻏﺘﻬﺎ ﰲ ﺷﻜﻞ ﳐﺘﺼﺮ‬

‫‪Z = W( b - b0 ) + U‬‬
‫ﺣﻴﻨﻬﺎ ﺗﺘﻢ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻘﺪﻳﺮ ﻟﺸﻌﺎﻉ ﺍﻟﺜﻮﺍﺑﺖ‬
‫^‬
‫‪( b - b 0 ) = ( W ' W ) -1 W ' Z‬‬ ‫)‪(5.44‬‬

‫ﺑﺈﻋﺎﺩﺓ ﺍﻟﺘﻌﻮﻳﺾ‬
‫^‬
‫]) ‪( b - b0 ) = [( G'( x, b0 )G( x, b0 )] -1 G'( x, b0 )[ Y - F( x, b0‬‬

‫‪188‬‬
 ‫ﺇﺫﺍﹶ‬
‫^‬
‫]) ‪b 1 = b 0 + [(G `( x, b 0 )G ( x, b 0 )] -1 G`( x, b 0 )[Y - F ( x, b 0‬‬
‫^‬
‫ﻫﻮ ﻣﻘﺪﺭ ﺍﻟﺘﻜﺮﺍﺭ ﺍﻷﻭﻝ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺷﻌﺎﻉ ﻗﻴ‪‬ﻢ ﺍﻻﻧﻄﻼﻕ ‪ b 0‬ﺍﻟﺬﻱ ﳝﻜﻦ‬
‫ﺍﳊﺼﻮﻝ ﻋﻠﻴﻪ ﻣﻦ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻣﺜﻼﹰ‪ .‬ﻭﻣﻦ ﻫﻨﺎ ﺗﺒﺪﺃ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻜﺮﺍﺭ ﺇﺫﺍ‬
‫^‬ ‫^‬
‫ﻛﺎﻧﺖ ﺣﺼﻴﻠﺔ ‪ b 1 - b 0‬ﲣﺘﻠﻒ ﻋﻦ ﺍﻟﺼﻔﺮ‬
‫‪16‬‬

‫^‬ ‫^‬
‫‪b1 - b0 ¹ 0‬‬
‫ﺣﻴﺚ‬
‫^‬ ‫^‬
‫]) ‪b 2 = b1 + [(G ¢( x , bˆ1 )G ( x, bˆ1 )]-1 G ¢( x, bˆ1 )[Y - F ( x, bˆ1‬‬

‫^‬ ‫^‬
‫‪ (b i - 1 - b i - 2‬ﻓﺈﻥ‬ ‫‪¹0‬‬ ‫ﻣﻨﻪ ﻭﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ )ﻭﺇﺫﺍ ﻛﺎﻧﺖ‬
‫^‬ ‫^‬
‫])‪bi = bi -1 + [(G¢( x, bˆi -1 )G(x, bˆi -1 )]-1G¢(x, bˆi -1 )[Y - F( x, bˆi -1‬‬

‫ﻫﻜﺬﺍ ﺗﺴﺘﻤﺮ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻜﺮﺍﺭ ﺣﱴ ﻳﻘﺘﺮﺏ‬

‫‪17‬‬

‫^‬ ‫^‬
‫‪b i - b i -1 » 0‬‬
‫^‬
‫ﻫﻮ ﺍﻟﺘﻘﺪﻳﺮ ﺍﻟﻨﻬﺎﺋﻲ‬ ‫‪bi‬‬ ‫ﺃﻳﻦ ﺗﺘﻮﻗﻒ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻜﺮﺍﺭ‪ ،‬ﻭﻳﻜﻮﻥ ﺍﻟﺸﻌﺎﻉ‬
‫ﳌﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ ﻏﲑ ﺍﳋﻄﻲ‪.‬‬

‫‪bˆi - bˆi -1‬‬

‫ﳛﺪﺩ‬ ‫‪d‬‬ ‫ﺣﻴﺚ‬ ‫‪bˆi‬‬
‫‪< d‬‬ ‫‪ b i - b i - 1 < d‬أو‬ ‫ﻣﻦ ﺍﳌﻌﺎﻳﲑ ﺍﳌﺴﺘﺨﺪﻣﺔ ﰲ ﻫﺬﺍ ﺍ‪‬ﺎﻝ ‪:‬‬ ‫‪- 16‬‬

‫ﻣﺴﺒﻘﺎﹰ‪.‬‬
‫‪ - 17‬ﺇﺫﺍ ﻛﺎﻥ ﻣﻦ ﻏﲑ ﺍﳌﻤﻜﻦ ﲢﻘﻴﻖ ﻫﺬﺍ ﺍﻟﺸﺮﻁ‪ ،‬ﻳﺘﻢ ﲢﺪﻳﺪ ﺳﻘﻒ ﳍﺬﻩ ﺍﻟﺘﻜﺮﺍﺭﺍﺕ ﻻ ﳝﻜﻦ‬
‫ﲡﺎﻭﺯﻩ‪.‬‬

‫‪189‬‬
‫ﻗﺪ ﺗﻮﺍﺟﻬﻨﺎ ﻋﻨﺪ ﺗﻜﺮﺍﺭ ﻣﻌﲔ ﺍﺣﺘﻤﺎﻝ ﻋﺪﻡ ﺇﻣﻜﺎﻧﻴﺔ ﻗﻠﺐ ﺍﳌﺼﻔﻮﻓﺔ‬
‫]) ˆ‪ ، [( G ¢( x , bˆ )G ( x , b‬ﺃﻳﻦ ﻧﻀﻄﺮ ﻹﺿﺎﻓﺔ ﺍﳌﻘﺪﺍﺭ ‪ l I‬ﺇﻟﻴﻬﺎ ﻣﻊ ‪ l‬ﻏﲑ‬ ‫‪i -1‬‬ ‫‪i -1‬‬

‫ﺻﻔﺮﻱ ﻣﻨﺎﺳﺐ ﻳﺴﻤﺢ ﺑﻘﻠﺐ ﺍﳌﺼﻔﻮﻓﺔ ﺃﻱ‪:‬‬

‫) ‪bˆi = bˆi -1 + [(G `(.) G (.) + l I ]- 1G `(.)[ Y - F ( x, bˆi -1‬‬

‫ﻧﺸﲑ ﺇﱃ ﺃ‪‬ﺎ‪ -‬ﻛﻐﲑﻫﺎ ﻣﻦ ﺍﻟﻄﺮﻕ ﺍﻟﺘﻜﺮﺍﺭﻳﺔ – ﺗﻌﺘﻤﺪ ﺑﺸﻜﻞ ﻛﺒﲑ ﻋﻨﺪ‬

‫ﺍﻟﺒﺤﺚ ﻋﻦ ﺍﳊﻞ‪ ،‬ﻋﻠﻰ ﺍﻟﻘﻴ‪‬ﻢ ﺍﻻﺑﺘﺪﺍﺋﻴﺔ ﺍﳌﺴﺘﻌﻤﻠﺔ‪ .‬ﻭﺍﻟﱵ ﻧﺬﻛﺮ ﺑﻌﺪﻡ ﻭﺟﻮﺩ ﻃﺮﻳﻘﺔ‬
‫ﻣﺜﻠﻰ ﺗﺴﻤﺢ ﺑﺎﻻﺧﺘﻴﺎﺭ ﺍﳌﻮﻓﻖ ﳍﺬﻩ ﺍﻟﻘﻴ‪‬ﻢ ﻟﻠﻮﺻﻮﻝ ﺇﱃ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ ﰲ‬
‫ﺃﻗﻞ ﻋﺪﺩ ﳑﻜﻦ ﻣﻦ ﺍﻟﺘﻜﺮﺍﺭﺍﺕ‪.‬‬
‫ﻫﺬﻩ ﺣﺎﻟﺔ ﻋﺎﻣﺔ ﻟﺘﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﳕﻮﺫﺝ ﻏﲑ ﺧﻄﻲ‪ ،‬ﻭﻟﺘﻄﺒﻴﻘﻬﺎ ﻋﻠﻰ‬
‫ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ‪-‬ﳏﻞ ﺩﺭﺍﺳﺘﻨﺎ‪ -‬ﻧﻘﻮﻡ ﺑﺎﻟﻌﻤﻠﻴﺎﺕ ﺍﻟﺘﺎﻟﻴﺔ ﺃﻳﻦ ﻳﻨﺤﺼﺮ ﻫﺪﻓﻨﺎ ﰲ‬
‫ﺗﺪﻧﻴﺔ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‪.‬‬

‫‪e t = Q -1 ( L ) F ( L ) y t‬‬

‫) ‪e t = f ( f, q, y t ) = f ( b , y t‬‬ ‫ﻣﻨﻪ‬

‫ﻟﻨﺸﺮ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﺣﻮﻝ ﻗﻴﻤﺔ ﺍﺑﺘﺪﺍﺋﻴﺔ ﻟﻠﺸﻌﺎﻉ ‪ b‬ﺍﳌﻤﺜﻞ ﻟﻠﻤﻌﻠﻤﺘﲔ ‪ f‬ﻭ‬
‫‪ q‬ﻭﺍﻟﱵ ﻧﺴﻤﻴﻬﺎ ‪ b 0‬ﻭﻓﻖ ﻣﻨﺸﻮﺭ ﺗﻴﻠﺮ ﳓﺼﻞ ﻋﻠﻰ‪:‬‬
‫) ‪¶f ( y t , b‬‬
‫‪b0‬‬
‫‪e t = f ( y t ,b0 ) = f ( y t ,b0 ) +‬‬ ‫‪( b - b 0 ) + rn‬‬
‫‪¶b‬‬

‫ﺍﻟﺸﻜﻞ)‪:(Matrix form‬‬ ‫ﺑﺈﳘﺎﻝ ﺍﻟﻌﻨﺼﺮ ﺍﳌﺘﺒﻘﻲ‪ ،‬ﻭﺑﻜﺘﺎﺑﺔ ﻣﺼﻔﻮﻓﺎﺗﻴﺔ‬

‫) ‪e t = e t /b 0 + G ( y t , b 0 ) ( b - b 0‬‬

‫‪190‬‬
‫ﺣﻴﺚ ﺍﳌﻘﺪﺍﺭ ‪ e t /b‬ﳝﺜﻞ ﺷﻌﺎﻉ ﺍﻟﺒﻮﺍﻗﻲ ﻣﻘﻮ‪‬ﻡ ﺑﻘﻴ‪‬ﻢ ﻣﻌﻠﻤﺎﺕ ﺍﻟﺸﻌﺎﻉ ‪b‬‬ ‫‪0‬‬

‫ﺍﻻﺑﺘﺪﺍﺋﻴﺔ ‪ ،b 0‬ﺇﺫﺍﹰ‬
‫‪e t / b 0 = - G ( y t , b 0 )( b - b 0 ) + e t‬‬ ‫)‪(5.45‬‬

‫ﻭﺑﻨﻔﺲ ﺍﻟﻄﺮﻳﻘﺔ‬
‫^‬
‫) ‪( b 0 - b ) = [ ( G '( y t , b 0 ) G( y t , b 0 ) ] -1 G '( y t , b 0 )( e t / b 0‬‬

‫ﺑﺸﻜﻞ ﺗﻜﺮﺍﺭﻱ‬
‫^‬
‫) ‪b1 = b0 - [(G¢( yt , b0 )G( yt , b0 )]-1 G¢( yt , b0 )(e t / b 0‬‬

‫ﻫﻮ ﺍﻟﺘﻜﺮﺍﺭ ﺍﻷﻭﻝ ﺍﻟﺬﻱ ﻳﻘﺪﺭ ‪ b 1‬ﲟﻌﻠﻮﻣﺔ ‪ ، b 0‬ﻭﺗﺴﺘﻤﺮ ﺍﻟﻌﻤﻠﻴﺔ‬

‫^‬ ‫^‬
‫ﺍﻟﺘﻜﺮﺍﺭﻳﺔ ﺣﱴ ﻳﺴﺠﻞ ﺍﺳﺘﻘﺮﺍﺭ ﰲ ﻗﻴﻤﺔ ﺷﻌﺎﻉ ﺍﳌﻌﻠﻤﺎﺕ‪ ،‬ﺃﻱ‪b i - b i -1 » 0 :‬‬
‫ﻋﻨﺪ ﺍﻟﺘﻜﺮﺍﺭ ‪i‬‬
‫ﻟﺘﻮﺿﻴﺢ ﺃﻛﺜﺮ ﳍﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﻧﺪﺭﺝ ﺍﳌﺜﺎﻝ ﺍﻟﺘﺎﱄ ﻟﺘﻘﺪﻳﺮ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﺨﺘﻠﻂ‬
‫)‪.ARMA(1.1‬‬
‫‪y t = fy t - 1 + e t + qe t - 1‬‬
‫ﺑﺈﺩﺧﺎﻝ ﻣﻌﺎﻣﻞ ﺍﻟﺘﺄﺧﲑ ‪L‬‬
‫‪y t - fLy‬‬ ‫‪t‬‬ ‫‪= e t + qL e t‬‬

‫‪( 1 - fL ) y t = ( 1 + fL ) e t‬‬

‫) ‪( 1 - fL‬‬
‫= ‪et‬‬ ‫‪yt‬‬ ‫ﻣﻨﻪ‬
‫) ‪( 1 + qL‬‬

‫ﺃﻳﻦ ﺗﻈﻬﺮ ﻫﺬﻩ ﺍﻟﻌﻼﻗﺔ ﰲ ﺷﻜﻞ ﻏﲑ ﺧﻄﻲ ﺍﳌﻌﻠﻤﺎﺕ‪ ،‬ﻭﺑﺎﺳﺘﻌﻤﺎﻝ ﻣﻨﺸﻮﺭ‬

‫ﺗﻴﻠﺮ‪:‬‬
‫‪e t = e t /( f 0,q0 ) + ¶e t /¶f0 ( f - f 0 ) + ¶e t /¶q0 ( q - q 0 ) + rn‬‬

‫‪191‬‬
 ‫ﺑﺈﻋﺎﺩﺓ ﻛﺘﺎﺑﺘﻬﺎ ﰲ ﺷﻜﻞ ﻣﺼﻔﻮﻓﺎﰐ‬

‫[‬
‫‪e t = e t /(f0 ,q 0 ) + ¶e t / ¶f0‬‬ ‫]‬‫‪éf - f 0 ù‬‬
‫‪¶e t / ¶q0 ê‬‬ ‫‪ú + rn‬‬
‫‪ëq - q 0 û‬‬

‫ﺑﺈﳘﺎﻝ ﺍﻟﻌﻨﺼﺮ ﺍﳌﺘﺒﻘﻲ‪ ،‬ﻧﻌﻴﺪ ﻛﺘﺎﺑﺘﻬﺎ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺴﺎﺑﻖ‬

‫) ‪e t = e t /b0 + G( y t , b 0 ) ( b - b 0‬‬

‫‪G ( y t , b 0 ) = ¶e t /¶f 0‬‬ ‫‪¶e t /¶q 0‬‬ ‫ﺣﻴﺚ‬

‫ﺃﻳﻦ ﳛﻮﻱ ﺷﻌﺎﻉ ﺍﳌﻌﻠﻤﺎﺕ ﰲ ﺣﺎﻟﺔ )‪ ARMA(1,1‬ﻣﻌﻠﻤﺘﲔ ﻓﻘﻂ ﺑﺎﻓﺘﺮﺍﺽ‬

‫ﻏﻴﺎﺏ ﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ‪.‬‬
‫‪éf ù‬‬
‫‪b0 = ê 0 ú‬‬
‫‪ëq 0 û‬‬
‫ﻛﻮﻥ )‪ G(.‬ﻫﻲ ﻣﺼﻔﻮﻓﺔ ﺍﳌﺸﺘﻘﺎﺕ ﺍﻷﻭﱃ ﻟﻠﺪﺍﻟﺔ ‪ f‬ﺑﺎﻟﻨﺴﺒﺔ ﳌﻌﻠﻤﺎﺕ‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﻗﻴﺪ ﺍﻟﺘﻘﺪﻳﺮ‪ ،‬ﺇﺫﺍﹰ‬
‫‪¶e t‬‬ ‫‪-L‬‬
‫=‬ ‫‪y t = g 1t‬‬ ‫)‪(5.46‬‬
‫‪¶f‬‬ ‫‪1 + qL‬‬

‫‪¶e t‬‬ ‫‪( 1 - fL ) L‬‬

‫‪=-‬‬ ‫‪y t = g 2t‬‬ ‫)‪(5.47‬‬
‫‪¶q‬‬ ‫‪( 1 + qL ) 2‬‬

‫ﻣﻦ ﺍﳌﺸﺘﻘﺘﲔ ﺃﻋﻼﻩ )‪ (5.46‬ﻭ)‪ (5.47‬ﻧﺴﺘﻨﺘﺞ ﻭﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪:‬‬

‫‪g 1 t + qg 1 t - 1 = - y t - 1‬‬

‫‪g 1 t = - y t - 1 - qg 1 t - 1‬‬

‫‪192‬‬
 ‫ﺑﺎﻟﺘﺸﺎﺑﻪ‬
‫‪( 1 - fL ) L‬‬
‫‪g 2t = -‬‬ ‫‪yt‬‬
‫‪( 1 + qL ) 2‬‬

‫‪g 2 t = - 2 q g 2 t - 1 - q 2 g 2 t - 2 - y t - 1 + fy t - 2‬‬

‫ﻣﻦ ﳘﺎ ﻧﻜﻮ‪‬ﻥ ﺍﳌﺼﻔﻮﻓﺔ‬

‫‪18‬‬

‫‪G( f, q, y t ) = G( b, y t ) = g 1 t‬‬ ‫‪g 2t‬‬

‫ﺑﻌﺪ ﺗﻘﻴ‪‬ﻴﻢ ﻫﺬﻩ ﺍﳌﺼﻔﻮﻓﺔ ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﻟﻘﻴ‪‬ﻢ ﺍﻻﺑﺘﺪﺍﺋﻴﺔ ﻟﻠﻤﻌﻠﻤﺎﺕ ﻭﻛﺬﺍ‬

‫ﻟﻠﻤﺘﻐﲑﺍﺕ ‪-‬ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﰲ ﻃﺮﻳﻘﺔ ﺍﳌﻌﻘﻮﻟﻴﺔ ﺍﻟﻌﻈﻤﻰ‪ -‬ﺗﺒﺪﺃ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻜﺮﺍﺭ‬
‫ﻭﻛﺎﻵﰐ‪:‬‬
‫^‬
‫) ‪b1 = b 0 - [(G ¢( yt , b 0 )G ( yt , b 0 )]-1 G ¢( yt , b 0 )(e t / b 0‬‬

‫ﺃﻳﻦ ﻭﰲ ﺍﻟﺘﻜﺮﺍﺭ ﺍﻷﻭﻝ‪:‬‬

‫‪g 1 t = - y t -1 - q 0 g 1 t -1‬‬
‫‪2‬‬
‫‪g2t = -2q0 g2t -1 - q0 g2t - 2 + yt -1 - f0 yt -2‬‬
‫ﺑﻨﻔﺲ ﺍﳌﻨﻬﺠﻴﺔ‬
‫‪e t /b 0 = y t - f 0 y t -1 - q 0 e t -1‬‬
‫ﻋﻨﺪ ﺍﻟﺘﻜﺮﺍﺭ ‪i‬‬
‫^‬ ‫^‬
‫) ‪bi = bi -1 - [(G¢( yt , bˆi -1)G( yt , bˆi -1)]-1G¢( yt , bˆi -1)(et /b i -1‬‬

‫‪ -18‬ﻭﻋﻠﻰ ﻏﺮﺍﺭ ﻃﺮﻕ ﺍﻟﺘﻤﻬﻴﺪ ﳓﺘﺎﺝ ﺇﱃ ﻗﻴ‪‬ﻢ ﺍﺑﺘﺪﺍﺋﻴﺔ ﻟﻠﻤﺘﻐﲑﺍﺕ ﺍﻟﺘﺎﺭﳜﻴﺔ ﻟﺘﺸﻜﻴﻞ ﻫﺬﻩ‬
‫ﺍﳌﺼﻔﻮﻓﺔ ) ‪. ( y 0 , y - 1 ,... , g 1, 0 , g 1 , - 1 ,... , g 2, 0 , g 2, - 1 ,..‬‬

‫‪193‬‬
 ‫ﺃﻳﻦ‬
‫‪g 1 t = - y t -1 - q i g 1 t - 1‬‬
‫‪2‬‬
‫‪g2t = -2qi g2t -1 - qi g2t - 2 + yt -1 - fi yt - 2‬‬

‫‪e t /b‬‬ ‫‪i -1‬‬

‫‪= y t - f i - 1 y t - 1 - q i -1 e t - 1‬‬ ‫ﻭ‬

‫ﺗﺘﻮﻗﻒ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻘﺪﻳﺮ ﺑﻨﻔﺲ ﺍﻟﺸﺮﻁ ﺍﳌﺬﻛﻮﺭ ﺳﺎﺑﻘﺎﹰ ﺣﻴﺚ ‪ i‬ﳝﺜﻞ ﺍﻟﺘﻜﺮﺍﺭ‬
‫ﺍﻷﺧﲑ ﺃﻱ‬
‫^‬ ‫^‬
‫‪.b i - b i -1 » 0‬‬
‫ﻭﺗﻘﺪﻳﺮ ﺍﻟﻨﻤﻮﺫﺝ ﺍﶈﺪﺩ ﺑﺎﺳﺘﻌﻤﺎﻝ‬ ‫‪GB13‬‬ ‫ﺑﺎﻟﺮﺟﻮﻉ ﺇﱃ ﻣﺜﺎﻝ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﺩﺍﻟﱵ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻜﻠﻴﺔ ﻭﺍﳉﺰﺋﻴﺔ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪yt = d + f1 yt -1 + f8 yt-8 + f12 yt -12 + q1e t-1 + q2e t -2 + q3e t-3 +q12e t-12‬‬

‫ﳓﺼﻞ ﻋﻠﻰ ﺍﻟﻨﺘﺎﺋﺞ ﺍﻟﺘﺎﻟﻴﺔ ﺑﻌﺪ )‪ (12‬ﺗﻜﺮﺍﺭ‪‬ﺍ ‪:‬‬

‫‪19‬‬

‫‪y t = 792283 . 5 + 0. 2 y t - 1 - 0. 1 y t - 8 + 0. 8 y t - 12 - 0. 01 e t - 1‬‬

‫) ‪( 550682‬‬ ‫) ‪( 0. 1‬‬ ‫) ‪( 0. 09‬‬ ‫) ‪( 0. 1‬‬ ‫) ‪( 0. 11‬‬

‫‪+ 0. 02 e t - 2 + 0. 007 e t - 3 - 0. 9 e t - 12‬‬

‫) ‪( 0. 11‬‬ ‫) ‪( 0. 1‬‬ ‫) ‪( 0. 1‬‬

‫‪R‬‬ ‫‪2‬‬
‫‪= 0. 66‬‬ ‫‪S = 65033 . 83‬‬ ‫‪T = 50‬‬

‫ﻧﺘﺮﻙ ﺗﻘﻴﻴﻢ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﳌﺎ ﺑﻌﺪ ﺗﻨﺎﻭﻝ ﻣﻮﺿﻮﻉ ﺍﻟﺘﺸﺨﻴﺺ‪.‬‬

‫‪20‬‬

‫‪ - 19‬ﻣﺎﺑﲔ ﺍﻟﻘﻮﺳﲔ ﺃﺩﻧﺎﻩ ﳝﺜﻞ ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ ﻟﻠﻤﻌﻠﻤﺔ‪ ،‬ﺑﻴﻨﻤﺎ ‪ S‬ﲤﺜﻞ ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ‬
‫ﻟﻼﳓﺪﺍﺭ‪ T ،‬ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ‪.‬‬

‫‪194‬‬
 ‫‪Diagnostic Checking‬‬ ‫‪ -3‬ﺗﺸﺨﻴﺺ ﺍﻟﻨﻤﻮﺫﺝ‪:‬‬
‫ﺑﻌﺪ ﺍﻻﻧﺘﻬﺎﺀ ﻣﻦ ﻣﺮﺣﻠﱵ ﲢﺪﻳﺪ ﻭﺗﻘﺪﻳﺮ ﺍﻟﻨﻤﻮﺫﺝ‪ ،‬ﻧﻮﺩ ﺍﻟﺘﻄﺮﻕ ﺇﱃ ﺍﳌﺮﺣﻠﺔ‬
‫ﺍﻟﺜﺎﻟﺜﺔ ﻣﻦ ﻋﻤﻠﻴﺔ ﺍﻟﻨﻤﺬﺟﺔ‪ ،‬ﻭﻫﻲ ﺍﺧﺘﺒﺎﺭ ﻗﻮﺓ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻹﺣﺼﺎﺋﻴﺔ ﰒ ﺍﻟﺘﻨﺒﺆﻳﺔ ﰲ‬
‫ﻣﺮﺣﻠﺔ ﻻﺣﻘﺔ‪ ،‬ﻭﻫﺬﻩ ﺍﳌﺮﺣﻠﺔ ﺗﺘﻄﻠﺐ ﺍﳋﻄﻮﺍﺕ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫ﺍ ‪ -‬ﻣﻘﺎﺭﻧﺔ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ ﻣﻊ ﺗﻠﻚ ﺍﳌﺘﻮﻟﺪﺓ ﻋﻦ‬
‫ﺍﻟﻨﻤﻮﺫﺝ )ﺍﳌﻘﺪﺭ(‪ ،‬ﻓﺈﺫﺍ ﻟﻮﺣﻆ ﻭﺟﻮﺩ ﺍﺧﺘﻼﻑ ﺟﻮﻫﺮﻱ ﺑﻴﻨﻬﻤﺎ‪ ،‬ﻳﻜﻮﻥ ﺫﻟﻚ‬
‫ﺩﻟﻴﻼﹰ ﻗﻄﻌﻴﺎﹰ ﻋﻠﻰ ﻓﺸﻞ ﻋﻤﻠﻴﺔ ﺍﻟﺘﺤﺪﻳﺪ‪ ،‬ﻭﻫﺬﺍ ﻳﺴﺘﺪﻋﻲ ﺇﻋﺎﺩﺓ ﻋﻤﻠﻴﺔ ﺑﻨﺎﺀ ﺍﻟﻨﻤﻮﺫﺝ‬
‫ﻭﺗﻘﺪﻳﺮﻩ ﻣﻦ ﺟﺪﻳﺪ‪.‬‬
‫ﺍﻟﺸﻜﻞ)‪ :(35‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﻨﻤﻮﺫﺝ ﺍﳌﻘﺪﺭ‬

‫ﺑـ ‪ -‬ﺃﻣﺎ ﺇﺫﺍ ﺗﺸﺎ‪‬ﺖ ﺍﻟﺪﺍﻟﺘﲔ ﻛﻤﺎ ﻫﻮ ﺍﳊﺎﻝ ﻋﻨﺪ ﻣﻘﺎﺭﻧﺔ ﺩﺍﻟﱵ ﺍﻟﺸﻜﻠﲔ‬
‫)‪ (32‬ﻭ)‪ (35‬ﺃﻋﻼﻩ‪ ،‬ﻧﻨﺘﻘﻞ ﺇﱃ ﺩﺭﺍﺳﺔ ﻭﲢﻠﻴﻞ ﺑﻮﺍﻗﻲ ﺍﻟﻨﻤﻮﺫﺝ‪ .‬ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ‬
‫ﺗﺘﻄﻠﺐ ﺣﺴﺎﺏ ﻭﺭﺳﻢ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﳍﺬﻩ ﺍﻟﺒﻮﺍﻗﻲ‪ ،‬ﰒﹼ ﺍﺧﺘﺒﺎﺭﻫﺎ ﻳﻜﻮﻥ ﻛﻤﺎ‬
‫ﻳﻠﻲ ‪:‬‬

‫‪ - 20‬ﻣﺎﺑﲔ ﺍﻟﻘﻮﺳﲔ ﳝﺜﻞ ﺩﺍﺋﻤﺎﹰ ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ ﻟﻠﻤﻌﻠﻤﺔ ﺍﳌﻘﺪﺭﺓ ﺃﻋﻼﻩ ‪.‬‬

‫‪195‬‬
‫ﺑـ ‪ -‬ﺍ ‪ -‬ﳚﺐ ﺃﻥ ﺗﻘﻊ ‪ -‬ﻛﻤﺎ ﻫﻮ ﻣﺒﲔ ﰲ ﺍﻟﺸﻜﻠﲔ ﺍﻟﺘﺎﻟﻴﲔ ‪ -‬ﻣﻌﻠﻤﺎﺕ‬
‫ﺩﺍﻟﱵ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻜﻠﻴﺔ ﻭﺍﳉﺰﺋﻴﺔ ﳍﺬﻩ ﺍﻟﺒﻮﺍﻗﻲ ﺩﺍﺧﻞ ﳎﺎﻝ ﺍﳌﻌﻨﻮﻳﺔ ﺍﳌﻌﱪ ﻋﻨﻪ‬
‫ﺑﻴﺎﻧﻴﺎﹰ ﲞﻄﲔ ﻣﺘﻮﺍﺯﻳﲔ ﺃﻭ ﺭﻳﺎﺿﻴﺎﹰ ﺑـ‪:‬‬
‫‪2‬‬
‫‪rk £‬‬
‫‪T‬‬

‫ﺍﻟﺸﻜﻞ )‪ :(36‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﺒﻮﺍﻗﻲ‬

‫ﺍﻟﺸﻜﻞ )‪ :(37‬ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ ﻟﻠﺒﻮﺍﻗﻲ‬

‫‪196‬‬
‫ﺑـ ‪ - 2 -‬ﻭﲢﺖ ﻓﺮﺿﻴﺔ ﺗﻮﺯﻳﻊ ﻣﻌﺎﻣﻼﺕ ﻫﺬﻩ ﺍﻟﺪﺍﻟﺔ ﻃﺒﻴﻌﻴﺎﹰ ﻭﺑﻮﺳﻂ ﻣﻌﺪﻭﻡ‬
‫ﻭﺗﺒﺎﻳﻦ ‪ 1‬ﺃﻱ ) ‪ r k ® N ( 0, 1‬ﻓﺈﻥ‪:‬‬
‫‪T‬‬ ‫‪T‬‬
‫‪k‬‬
‫‪Q = T å ri2 ® c 2‬‬
‫‪i=1‬‬ ‫‪k - p- q‬‬

‫ﺍﻟﱵ ﺗﻌﺎﺩﻝ ﰲ ﻣﺜﺎﻟﻨﺎ ﻫﺬﺍ ‪ ،14.84‬ﺣﻴﺚ ‪ rk‬ﻣﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬

‫) ‪å ( e t .e t - k‬‬
‫= ‪rk‬‬ ‫‪2‬‬
‫‪å et‬‬
‫ﺃﻳﻦ ‪ et‬ﺗﺴﺎﻭﻱ‬
‫‪$ ( L )Q‬‬
‫‪e =F‬‬ ‫‪$ -1 ( L ) y‬‬
‫‪t‬‬ ‫‪t‬‬
‫ﲟﻘﺎﺭﻧﺔ ‪ Q‬ﻣﻊ ‪ c 28,10% = 20. 090‬ﻧﻘﺒﻞ ﻣﺒﺎﺷﺮﺓ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ‪ ،‬ﲟﻌﲎ ﺃﻥ‬
‫ﻛﻞ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻟﻠﺒﻮﺍﻗﻲ ﻣﻌﺪﻭﻣﺔ‪ ،‬ﻭﺃﺧﺬﻧﺎ ‪ a = 10%‬ﻛﻮﻥ‬
‫ﻣﺴﺘﻮﻯ ‪ 5%‬ﻣﻌﻨﻮﻳﺔ ﻻ ﻳﻔﻲ ﺑﺎﻟﻐﺮﺽ ﰲ ﺃﺣﻮﺍﻝ ﻛﺜﲑﺓ ‪.‬‬
‫ﺑﺪﻝ ‪ Q‬ﺍﻟﺬﻱ ﻳﻌﺎﱐ ﻣﻦ ﻧﻔﺲ ﻋﻴﺐ‬ ‫*‪Q‬‬ ‫ﻧﺸﲑ ﻫﻨﺎ ﺇﱃ ﺇﻣﻜﺎﻧﻴﺔ ﺍﺳﺘﻌﻤﺎﻝ‬
‫‪ R 2‬ﺍﳌﺬﻛﻮﺭ ﺳﺎﺑﻘﺎﹰ ﻭﻟﺬﺍ‬
‫‪k‬‬
‫‪Q * = T ( T + 2 ) å ( T - i ) - 1 ri ® c 2‬‬
‫‪2‬‬
‫‪k - p-q‬‬
‫‪i=1‬‬

‫‪Modified‬‬ ‫ﻳﻌﺮﻑ ﻫﺬﺍ ﺍﻷﺧﲑ ﺑﺎﺧﺘﺒﺎﺭ ‪ Ljung-Box-Pierce Statistic‬ﺃﻭ‬

‫‪Eviews‬‬ ‫‪ Box-Pierce‬ﻭﳛﺴﺐ ﻣﻦ ﻃﺮﻑ ﻣﻌﻈﻢ ﺍﻟﱪﺍﻣﺞ ﺍﻹﺣﺼﺎﺋﻴﺔ ﻣﺜﻞ‬
‫‪...‬ﺍﱁ‪.‬‬
‫ﺍﶈﺴﻮﺑﺔ ﻟﻸﺧﻄﺎﺀ ﺃﻗﻞ ﻣﻦ ﺗﻠﻚ ﺍ‪‬ﺪﻭﻟﺔ ﺣﱴ‬ ‫*‪Q‬‬ ‫ﳚﺐ ﺃﻥ ﺗﻜﻮﻥ ‪ Q‬ﻭ‬
‫ﻳﺘﻮﺍﻓﻖ ﺑـ‪ 1-‬ﻣﻊ ﺑـ‪ 2-‬ﺃﻋﻼﻩ‪.‬‬

‫‪197‬‬
‫ﺗﻜﻮﻥ ﰲ ﻣﺜﺎﻟﻨﺎ ‪ Q * = 17. 75‬ﻫﻮ ﻧﻔﺲ ﺍﻟﻘﺮﺍﺭ ﺍﳌﺘﻮﺻﻞ ﺇﻟﻴﻪ ﺃﻋﻼﻩ ﺑﻌﺪ‬
‫ﺭﻓﻊ ﻣﺴﺘﻮﻯ ﺍﳌﻌﻨﻮﻳﺔ ﺇﱃ ‪ 10%‬ﻭ ﻫﺬﺍ ﺍﻹﺟﺮﺍﺀ ﻭﺍﺭﺩ ﰲ ﻫﺬﻩ ﺍﻟﻈﺮﻭﻑ ﻧﻈﺮﺍﹰ‬
‫ﻟﻀﻌﻒ ﺍﳌﻌﻨﻮﻳﺔ ﰲ ﺍﳌﻴﺪﺍﻥ ﺍﻟﺘﻄﺒﻴﻘﻲ‪.‬‬
‫ﺟـ‪ -‬ﻭﻧﻈﺮﺍﹰ ﻟﻠﺘﻮﺯﻳﻊ ﺍﻟﻄﺒﻴﻌﻲ ﺍﻟﺘﻘﺎﺭﰊ ﳌﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﺑﻮﺳﻂ‬
‫ﻣﻌﺪﻭﻡ ﻭﺗﺒﺎﻳﻦ ﻣﻌﲔ‪ ،‬ﻓﺈﻥ ﺍﻹﺣﺼﺎﺀﺓ ‪ t‬ﻭ ‪ fisher‬ﺗﺼﺒﺢ ﻏﲑ ﻣﱪﺭﺓ ﺍﻻﺳﺘﻌﻤﺎﻝ‪،‬‬
‫ﻛﺒﺪﻳﻞ ﳍﺬﻩ‪ ،‬ﻧﺴﺘﻌﻤﻞ ﻛﻞ ﻣﻦ )‪ N(O,1‬ﻭ ‪ c 2‬ﻭﺍﻟﻠﺘﺎﻥ ﺗﺄﺧﺬﺍﻥ ﺍﻟﺸﻜﻠﲔ ﺍﻟﺘﺎﻟﻴﲔ‬
‫ﻣﺜﻼﹰ‪:‬‬
‫^‬
‫‪b -b‬‬
‫‪j‬‬ ‫‪j‬‬
‫) ‪® N( 0, 1‬‬
‫^‬
‫) ‪se( b‬‬
‫‪j‬‬

‫ﻫﺬﺍ ﺍﳌﻌﻴﺎﺭ ﺧﺎﺹ ﺑﺎﺧﺘﺒﺎﺭ ﺍﳌﻌﻠﻤﺔ ‪ b‬ﺣﻴﺚ ‪ b j‬ﻗﻴﻤﺔ ﺍﳌﻌﻠﻤﺔ ﲢﺖ ﻓﺮﺿﻴﺔ‬

‫^‬
‫‪j‬‬

‫ﺍﻟﻌﺪﻡ‬
‫^‬
‫‪H :b = 0‬‬ ‫‪, vs H : b ¹ 0‬‬
‫‪0 j‬‬ ‫‪a j‬‬
‫^‬
‫ﺑﻴﻨﻤﺎ ) ‪ se ( b J‬ﻋﺒﺎﺭﺓ ﻋﻦ ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ ﻟﻠﻤﻌﻠﻤﺔ ﺍﳌﻌﻨﻴﺔ ‪-‬ﺑﲔ‬
‫ﺍﻟﻘﻮﺳﲔ‪-‬ﻫﻨﺎ‪ .‬ﻛﻤﺎ ﺃﻥ ﺍﺧﺘﺒﺎﺭ ﳎﻤﻮﻋﺔ ﻣﻦ ﺍﳌﻌﻠﻤﺎﺕ ﺁﻧﻴﺎﹰ ﻻ ﻳﺘﻢ ﺑﺪﻻﻟﺔ‬
‫ﺇﺣﺼﺎﺀﺓ ‪ fisher‬ﺇﳕﺎ ﺑﻮﺍﺳﻄﺔ ﺍﻹﺣﺼﺎﺀﺓ ‪ c 2‬ﺍﳌﻌﻄﺎﺓ ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫)‪(RRSS - URSS‬‬
‫‪®c2‬‬
‫‪URSS‬‬ ‫‪m‬‬
‫‪T‬‬

‫ﻫﻲ‬ ‫ﻭ‪ RRSS‬ﻭ‪URSS‬‬ ‫ﺣﻴﺚ ‪ m‬ﻫﻲ ﻋﺪﺩ ﺍﳌﻌﻠﻤﺎﺕ ﺍﳌﺮﺍﺩ ﺍﺧﺘﺒﺎﺭﻫﺎ‪،‬‬

‫ﻋﺒﺎﺭﺓ ﻋﻦ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ﲢﺖ ﺍﻟﻔﺮﺿﻴﺔ ‪ H 0‬ﻭ‪ H a‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪.‬‬

‫‪198‬‬
‫ﺩ‪ -‬ﻧﺸﲑ ﻫﻨﺎ ﻹﻣﻜﺎﻧﻴﺔ ﲡﺎﻭﺯ ﺑﻌﺾ ﺍﻟﻨﻤﺎﺫﺝ ﳍﺬﻩ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ‪ ،‬ﻭﻟﻠﻘﻴﺎﻡ‬
‫ﺑﻌﻤﻠﻴﺔ ﺍﳌﻔﺎﺿﻠﺔ ﺑﻴﻨﻬﺎ )ﺍﻟﻨﻤﺎﺫﺝ ﻏﲑ ﺍﳌﺮﻓﻮﺿﺔ ﺑﻮﺍﺳﻄﺔ ﺍﻷﺩﻭﺍﺕ ﺍﻹﺣﺼﺎﺋﻴﺔ ﺍﻟﺴﺎﺑﻘﺔ‬
‫ﺍﻟﺬﻛﺮ( ﻧﺴﺘﻌﻤﻞ ﺍﳌﻘﺎﻳﻴﺲ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫‪21‬‬

‫^‬ ‫‪p+ q‬‬

‫( ‪AIC = s 2 . exp{ 2‬‬ ‫})‬
‫‪T‬‬
‫^‬
‫ﺣﻴﺚ ‪ s 2 = S 2‬ﳏﺴﻮﺑﺎﹰ ﺑﻄﺮﻳﻘﺔ ﺍﳌﻌﻘﻮﻟﻴﺔ ﺍﻟﻌﻈﻤﻰ ﺃﻱ ﺑﻘﺴﻤﺔ ﳎﻤﻮﻉ‬
‫ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ﻋﻠﻰ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ﻓﻘﻂ ﻛﻤﺎ ﺃﻥ ﺍﳌﻘﺪﺍﺭ ) ‪ ( p + q‬ﻫﻨﺎ ﻳﺸﲑ‬
‫ﺇﱃ ﻋﺪﺩ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻘﺪﺭ ﻭﻟﻴﺲ ﳎﻤﻮﻉ ﺩﺭﺟﱵ ﺍﻟﻨﻤﻮﺫﺝ‪ .‬ﳝﻜﻦ ﻛﺘﺎﺑﺔ‬
‫ﻫﺬﺍ ﺍﳌﻌﻴﺎﺭ ﰲ ﺷﻜﻠﻪ ﺍﻟﻠﻮﻏﺎﺭﲤﻲ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫^‬ ‫‪p+q‬‬
‫( ‪AIC = ln ( s 2 ) + { 2‬‬ ‫})‬
‫‪T‬‬

‫ﺑﺴﺒﺐ ﺇﻋﻄﺎﺋﻪ ﻭﺯﻥ ﺃﻛﱪ ﻟﻠﻨﻤﺎﺫﺝ ﺍﳌﺴﺘﻌﻤ‪‬ﻠﺔ ﻷﻛﱪ ﻋﺪﺩ ﻣﻦ ﺍﳌﺸﺎﻫﺪﺍﺕ‬

‫ﻋ‪‬ﺪ‪‬ﻝ ﺑﺎﺧﺘﺒﺎﺭ )‪ (NAIC‬ﺍﻟﺘﺎﱄ ‪:‬‬
‫‪22‬‬

‫‪AIC‬‬
‫= ‪NAIC‬‬
‫‪T‬‬
‫ﺃﻳﻦ ﻳﻜﻮﻥ ﺍﻻﺧﺘﻴﺎﺭ ﻋﻠﻰ ﺃﺳﺎﺱ ﺃﺻﻐﺮ ﻗﻴﻤﺔ ﻟﻠﻤﻌﻴﺎﺭ‪ ،‬ﺃﻱ ﻧﻔﻀﻞ ﺍﻟﻨﻤﻮﺫﺝ‬
‫ﺍﻟﺬﻱ ﳛﻘﻖ ﺃﺻﻐﺮ ‪ AIC‬ﺃﻭ ‪.NAIC‬‬
‫‪SHWARZ‬‬ ‫ﺭﻏﺒﺔﹰ ﰲ ﲢﻘﻴﻖ ﺧﺼﺎﺋﺺ ﺗﻘﺎﺭﺑﻴﺔ ﻭﻷﺳﺒﺎﺏ ﺃﺧﺮﻯ ﺍﻗﺘﺮﺡ‬
‫ﺍﻟﺘﻌﺪﻳﻞ ﺍﻟﺘﺎﱄ ‪:‬‬
‫‪23‬‬

‫^‬
‫)‪( p + q‬‬
‫‪BIC = ln s 2 +‬‬ ‫‪ln T‬‬
‫‪T‬‬

‫‪21- Akaike Information Criterion.‬‬

‫‪22 - Normalised Akaike Information Criterion.‬‬
‫‪23 - Bayesian Information Criterion.‬‬

‫‪199‬‬
‫ﻻﺳﺘﺒﻴﺎﻥ ﻛﻴﻔﻴﺔ ﺍﺳﺘﻌﻤﺎﻝ ﻫﺬﻩ ﺍﳌﻌﺎﻳﲑ‪ ،‬ﻧﺴﺘﺮﺟﻊ ﻣﺸﻜﻞ ‪ GB13‬ﻭﺍﻟﻨﻤﻮﺫﺝ‬
‫ﺍﳌﻘﺪﺭ ﻋﻨﺪ ﺩﺭﺍﺳﺔ ﻃﺮﻕ ﺍﻟﺘﻘﺪﻳﺮ‪ ،‬ﺍﻟﺬﻱ ﻧﻌﻴﺪ ﻃﺒﻌﻪ ﰲ ﺷﻜﻠﻪ ﺍﻷﺻﻠﻲ ﺻﻤﻦ‬
‫ﺍﳉﺪﻭﻝ )‪ (35‬ﺍﻟﺘﺎﱄ‪:‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(34‬ﻧﺘﺎﺋﺞ ﺍﻟﺘﻘﺪﻳﺮ ﻟﻨﻤﻮﺫﺝ ‪ GB13‬ﺑـ ‪ 50‬ﻣﺸﺎﻫﺪﺓ‬
‫ﺍﻻﳓﺮﺍﻑ ﺍﳌﻌﻴﺎﺭﻱ‬ ‫ﺍﳌﻌﺎﻣﻞ‬ ‫ﺍﳌﺘﻐﲑ‬
‫‪550681.7‬‬ ‫‪792283.51‬‬ ‫ﺍﻟﺜﺎﺑﺖ‬
‫‪0.1120‬‬ ‫‪-0.0111‬‬ ‫)‪MA(1‬‬
‫‪0.1109‬‬ ‫‪0.0220‬‬ ‫)‪MA(2‬‬
‫‪0.1002‬‬ ‫‪0.0076‬‬ ‫)‪MA(3‬‬
‫‪0.0627‬‬ ‫‪-0.9099‬‬ ‫)‪MA(12‬‬
‫‪0.1046‬‬ ‫‪0.2026‬‬ ‫)‪AR(1‬‬
‫‪0.0965‬‬ ‫‪-0.0944‬‬ ‫)‪AR(8‬‬
‫‪0.1037‬‬ ‫‪0.7934‬‬ ‫)‪AR(12‬‬

‫ﺍﺧﺘﺒﺎﺭ ﺍﳌﻌﻨﻮﻳﺔ‪:‬‬
‫ﻓﺒﻤﻘﺎﺭﻧﺔ ﺍﻹﺣﺼﺎﺀﺓ ﺍﶈﺴﻮﺑﺔ ‪ Ncal‬ﻣﻊ ﺍ‪‬ﺪﻭﻟﺔ ﻭﺍﳌﻌﺎﺩﻟﺔ ﻝ‪،1.96‬‬
‫‪24‬‬

‫ﻧﺴﺘﻨﺘﺞ ﺃﻥ ﺍﳌﻌﺎﻣﻼﺕ ﺍﻟﺘﺎﻟﻴﺔ‪ ،‬ﻏﲑ ﻣﻌﻨﻮﻳﺔ ﲤﺎﻣﺎﹰ ﻣﺎ ﻋﺪﺍ ﺗﻠﻚ ﺍﻟﻌﺎﻛﺴﺔ ﻟﻠﻔﺼﻠﻴﺔ ﻣﻊ‬
‫ﻗﺒﻮﻝ ﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ ﻭﻣﻌﻠﻤﺔ )‪ AR(1‬ﻟﻘﺮ‪‬ﺎ ﻣﻦ ﻣﺴﺘﻮﻯ ﺍﳌﻌﻨﻮﻳﺔ‪.‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(35‬ﻧﺘﺎﺋﺞ ﺍﻟﺘﻘﺪﻳﺮ ﻣﻊ ﺍﺧﺘﺒﺎﺭ ﺍﳌﻌﻨﻮﻳﺔ‬
‫‪N-‬‬
‫اﻟﻘﺮار‬ ‫اﻻﻧﺤﺮاف اﻟﻤﻌﯿﺎري‬ ‫اﻟﻤﻌﻠﻤﺔ‬ ‫اﻟﻤﺘﻐﯿﺮ‬
‫‪stat‬‬
‫ﻏﯿﺮ ﻣﻌﻨﻮي‬ ‫‪1.44‬‬ ‫‪550681.7‬‬ ‫‪792283‬‬ ‫‪C‬‬
‫ﻗﺮﯾﺐ ﻣﻦ اﻟﻤﻌﻨﻮﯾﺔ *‬ ‫‪1.93‬‬ ‫‪0.1‬‬ ‫‪0.2‬‬ ‫)‪AR(1‬‬
‫ﻏﯿﺮ ﻣﻌﻨﻮي‬ ‫‪-097‬‬ ‫‪0.09‬‬ ‫‪-0.09‬‬ ‫)‪AR(8‬‬
‫ﻣﻌﻨﻮي *‬ ‫‪7.64‬‬ ‫‪0.1‬‬ ‫‪0.79‬‬ ‫)‪AR(12‬‬
‫ﻏﯿﺮ ﻣﻌﻨﻮي‬ ‫‪-0.09‬‬ ‫‪0.11‬‬ ‫‪-0.01‬‬ ‫)‪MA(1‬‬
‫ﻏﯿﺮ ﻣﻌﻨﻮي‬ ‫‪0.19‬‬ ‫‪0.11‬‬ ‫‪0.02‬‬ ‫)‪MA(2‬‬
‫ﻏﯿﺮ ﻣﻌﻨﻮي‬ ‫‪0.07‬‬ ‫‪0.1‬‬ ‫‪0.007‬‬ ‫)‪MA(3‬‬
‫ﻣﻌﻨﻮي *‬ ‫‪-14.49‬‬ ‫‪0.06‬‬ ‫‪-0.9‬‬ ‫)‪MA(12‬‬

‫‪ - 24‬ﰲ ﺣﺎﻟﺔ ﺃﺧﺬ ‪ %5=a‬ﳝﺜﻞ ﻫﺬﺍ ﺍﻟﺮﻗﻢ )‪ N(0.1‬ﺍ‪‬ﺪﻭﻟﺔ‪.‬‬

‫‪200‬‬
‫ﻓﺈﺫﺍ ﺃﺭﺩﻧﺎ ﺍﺧﺘﺒﺎﺭ ﻣﻌﻨﻮﻳﺔ ﺍﳌﻌﻠﻤﺔ ﺍﻷﻭﱃ ﰲ ﻫﺬﺍ ﺍﳉﺪﻭﻝ‪ ،‬ﺗﻜﺘﺐ ﺻﻴﻐﺘﻪ‬
‫ﺍﻻﺧﺘﺒﺎﺭﻳﺔ ﻛﻤﺎ ﻳﻠﻲ ‪:‬‬
‫‪H 0:C = 0‬‬
‫‪H a:C ¹ 0‬‬
‫ﺣﻴﺚ ‪ N - stat‬ﻟﻠﺤﺪ ﺍﻟﺜﺎﺑﺖ ﺍﳌﻌﲏ ﻫﻨﺎ‪ ،‬ﳛﺴﺐ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪C‬‬‫‪$ -C‬‬ ‫‪792283. 51‬‬

‫=‬ ‫‪= 1. 438‬‬
‫‪$‬‬
‫) ‪se( C‬‬ ‫‪550681. 7‬‬
‫ﺃﻣﺎ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﻌﻠﻤﺔ ﺍﻷﺧﲑﺓ‬
‫‪q$ 12 - q 12‬‬ ‫‪-0. 9‬‬
‫=‬ ‫‪= -14. 49‬‬
‫) ‪se( q$‬‬ ‫‪12‬‬
‫‪0. 06‬‬

‫‪q$ 12 - q 12‬‬
‫‪= 14. 49‬‬ ‫ﻣﻨﻪ‬
‫) ‪se ( q$‬‬
‫‪12‬‬

‫ﻋﺪﺍ ﺍﻟﺘﺄﺛﲑﺍﺕ ﺍﻟﻔﺼﻠﻴﺔ‪ ،‬ﻳﻈﻬﺮ ﺟﻠﻴﺎﹰ ﺃﻥ ﻛﻞ ﺍﳌﻌﻠﻤﺎﺕ ﺍﻟﺒﺎﻗﻴﺔ ﻏﲑ ﻣﻌﻨﻮﻳﺔ‬

‫ﳍﺬﺍ ﻭﺑﻐﻴﺔ ﲢﺴﲔ ﺍﻟﻨﻤﻮﺫﺝ‪ ،‬ﻭﺑﺎﺗﺒﺎﻉ ﺃﺳﻠﻮﺏ ﺇﺯﺍﻟﺔ ﺍﳌﺘﻐﲑﺍﺕ ﻏﲑ ﺍﳌﻌﻨﻮﻳﺔ ﻛﻞ‬
‫ﻭﺍﺣﺪ ﻋﻠﻰ ﺣﺪﺍ ﻭﺑﺪﺍﻳﺔ ﻣﻦ ﺍﻷﺿﻌﻒ‪ ،‬ﰎ ﺍﻟﺘﻮﺻﻞ ﺇﱃ ﺍﻟﺒﺪﺍﺋﻞ ﺍﻟﺘﺎﻟﻴﺔ ﻭﺍﻋﺘﻤﺎﺩﺍ ﻋﻠﻰ‬
‫ﺍﳌﻌﻨﻮﻳﺔ ﻓﻘﻂ‬
‫‪25‬‬

‫)‪(1‬‬ ‫ﺍﻟﻨﻤﻮﺫﺝ‬
‫^‬
‫‪y t = 373151. 28 + 0. 66 y t - 1 + 0. 87e t - 12‬‬
‫} ‪{ 15. 16‬‬ ‫} ‪{ 8. 8‬‬ ‫} ‪{ 27. 73‬‬

‫‪R 2 = 0. 68‬‬ ‫‪S ml = 62193. 37‬‬ ‫‪T = 61‬‬

‫^‬
‫= ‪. S ml = s ml‬‬
‫‪RSS‬‬
‫‪T‬‬
‫‪ -25‬ﻣﺎ ﺑﲔ ﺍﻟﻘﻮﺳﲔ ﳝﺜﻞ ‪ N stat‬ﻟﻠﻤﻌﻠﻤﺔ ﺍﳌﻌﻨﻴﺔ ﺃﻋﻼﻩ ﻭ‬

‫‪201‬‬
 ‫)‪(2‬‬ ‫ﺍﻟﻨﻤﻮﺫﺝ‬
‫^‬
‫‪y t = 373151. 28 + 0. 36e t - 1 + 0. 58 e t - 12‬‬
‫} ‪{ 35. 29‬‬ ‫} ‪{ 3. 5‬‬ ‫} ‪{ 6. 71‬‬

‫‪R 2 = 0. 49 S ml = 79187. 4‬‬ ‫‪T = 62‬‬

‫)‪(3‬‬ ‫ﺍﻟﻨﻤﻮﺫﺝ‬
‫^‬
‫‪y t = 504948. 6 + 0. 44 y t - 1 + 0. 4 y t - 12‬‬
‫} ‪{ 3. 8‬‬ ‫} ‪{ 3. 95‬‬ ‫} ‪{ 3. 82‬‬

‫‪R 2 = 0. 52 S ml = 71215. 62‬‬ ‫‪T = 50‬‬

‫)‪(4‬‬ ‫ﺍﻟﻨﻤﻮﺫﺝ‬
‫^‬
‫‪y t = 0. 32 y t - 1 + 0. 798 y t - 12 - 0. 896 e t - 12‬‬
‫} ‪{ 4. 22‬‬ ‫} ‪{ 9. 23‬‬ ‫} ‪{ -16. 86‬‬

‫‪R 2 = 0. 62 S ml = 63225. 72‬‬ ‫‪T = 50‬‬

‫ﻛﻮﻥ ﺍﻟﻨﻤﻮﺫﺟﲔ ﺍﻷﺧﲑﻳﻦ ﻣﺘﺴﺎﻭﻳﻦ ﻣﻦ ﺣﻴﺚ ﻋﺪﺩ ﺍﳌﻌﻠﻤﺎﺕ‬

‫ﻭﺍﳌﺸﺎﻫﺪﺍﺕ ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‪ ،‬ﻓﻌﻨﺪ ﺍﳌﻔﺎﺿﻠﺔ ﺑﻴﻨﻬﻤﺎ ﺑﺎﺳﺘﻌﻤﺎﻝ ﻣﻘﻴﺎﺱ ‪ ،AIC‬ﳜﺘﺼﺮ‬
‫ﰲ ﺍﻻﺧﺘﻴﺎﺭ ﻟﻴﺼﺒﺢ ﻋﻠﻰ ﺃﺳﺎﺱ ﺃﺻﻐﺮ ﺍﳓﺮﺍﻑ ﻣﻌﻴﺎﺭﻱ ﻟﻼﳓﺪﺍﺭ) ‪ ، ( S‬ﻭﻣﻨﻪ‬
‫ﻓﻨﻠﻐﻲ ﺍﻟﺜﺎﻟﺚ ﻣﻦ ﻣﻴﺪﺍﻥ ﺍﳌﻔﺎﺿﻠﺔ ﻟﺘﺒﻘﻰ ﺍﻷﺧﺮﻯ ﻣﺪﺭﺟﺔ ﰲ ﺍﳉﺪﻭﻝ ﺍﻟﺘﺎﱄ‪:‬‬
‫ﺍﳉﺪﻭﻝ )‪ :(36‬ﻣﻔﺎﺿﻠﺔ ﺍﻟﻨﻤﺎﺫﺝ‬
‫‪NAIC‬‬ ‫‪AIC‬‬ ‫‪T‬‬ ‫ﺍﻟﻨﻤﻮﺫﺝ‬
‫‪0.3634‬‬ ‫‪22.17‬‬ ‫‪61‬‬ ‫‪1‬‬
‫‪0.3651‬‬ ‫‪22.64‬‬ ‫‪62‬‬ ‫‪2‬‬
‫‪0.4446‬‬ ‫‪22.23‬‬ ‫‪50‬‬ ‫‪4‬‬

‫‪202‬‬
‫ﳔﺘﺎﺭ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻷﻭﻝ ﻋﻠﻰ ﺃﺳﺎﺱ ‪ NAIC‬ﻛﻮﻥ ﻋﺪﺩ ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﳌﺴﺘﻌﻤﻠﺔ‬
‫ﺃﺛﻨﺎﺀ ﺍﻟﺘﻘﺪﻳﺮ ﻏﲑ ﻣﺘﺴﺎﻭﻳﺔ‪.‬ﻛﻤﺎ ﺃﻥ ﺍﻟﻘﺮﺍﺭ ﱂ ﳜﺘﻠﻒ ﰲ ﻫﺬﺍ ﺍﳌﺜﺎﻝ ﻋﻦ ﻧﺘﻴﺠﺔ‬
‫‪ .AIC‬ﻧﺘﺮﻙ ﻟﻠﻄﺎﻟﺐ ﻓﺮﺻﺔ ﺣﺴﺎﺏ ﻣﻌﻴﺎﺭ ‪ BIC‬ﻟﻠﺒﺪﺍﺋﻞ ﺍﻟﻮﺍﺭﺩﺓ ﺃﻋﻼﻩ ﻭﺍﳌﻔﺎﺿﻠﺔ‬
‫ﺑﻴﻨﻬﺎ‪.‬‬

‫‪ -4‬ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﺍﻟﺸﻌﺎﻋﻴﺔ ﻭ ﺍﻹﺳﺘﻘﺮﺍﺭﻳﺔ ‪:‬‬

‫ﳝﺜﻞ ﻫﺬﺍ ﺍﻟﻌﻤﻞ‪ ،‬ﻣﻘﺪﻣﺔ ﻟﻠﺘﻄﻮﺭﺍﺕ ﻭﺍﳌﺴﺘﺠﺪﺍﺕ ﺍﻟﻌﻠﻤﻴﺔ ﰲ ﻣﻴﺪﺍﻥ ﺑﻨﺎﺀ‬
‫ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺘﻨﺒﺆﻳﺔ ﺍﻟﱵ ﺍﻋﺘﻤﺪﺕ ﻣﺆﺧﺮﺍﹰ ﻭﺑﺪﺍﻳﺔ ﻣﻦ ﻋﻤﻞ ﻛﺮﻳﺴﺘﻮﻓﺮ ﺳﻴﻤﺲ ]‪[44‬‬
‫))‪ Christopher Sims(1980‬ﺍﻹﻧﺘﻘﺎﺩﻱ ﻟﻠﺠﻨﺔ ﻛﻮ‪‬ﻟﺰ )‪(Cowels Commission‬‬
‫ﺍﻟﱵ ﺍﻋﺘﻤﺪﺕ ﻋﻠﻰ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻵﻧﻴﺔ ﺍﳍﻴﻜﻠﻴﺔ ﰲ ﻋﻤﻠﻴﺎ‪‬ﺎ ﺍﻟﺘﻨﺒﺆﻳﺔ ﺍﳌﺒﻨﻴﺔ ﻋﻠﻰ‬
‫ﻓﺮﺿﻴﺎﺕ ﻛﺜﲑﺓ ‪-‬ﺑﻐﻴﺔ ﺍﻟﻮﺻﻮﻝ ﺇﱃ ﳕﻮﺫﺝ ﳏﺪﺩ )‪-(Model Identification‬‬
‫ﻏﲑ ﳐﺘﱪﺓ ﻭ ﱂ ﻳﺘﻢ ﺍﻟﺘﺄﻛﺪ ﻣﻦ ﺟﺪﻭﺍﻫﺎ ﳑﺎ ﺍﺳﺘﻮﺟﺐ ﺇﺑﻌﺎﺩ ﻣﺘﻐﲑﺍﺕ ﻛﺜﲑﺓ ﻣﻦ‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﺑﻄﺮﻳﻘﺔ ﺇﻗﺼﺎﺋﻴﺔ ﺃﻭ ﺍﻧﺘﻘﺎﺋﻴﺔ ﻻ ﻣﱪﺭﺓ ﳍﺎ‪ ،‬ﺳﻮﺍﺀ ﻛﺎﻧﺖ ﻫﺬﻩ ﺍﳌﺘﻐﲑﺍﺕ‬
‫ﺩﺍﺧﻠﻴﺔ ﻣﺆﺧﺮﺓ ﺃﻭ ﺧﺎﺭﺟﻴﺔ‪ .‬ﻛﺒﺪﻳﻞ ﳍﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ‪ ،‬ﺍﻗﺘﺮﺡ ﻛﺮﻳﺴﺘﻮﻓﺮ ﺳﻴﻤﺲ‬
‫ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﺍﻟﺸﻌﺎﻋﻴﺔ‪ ،26‬ﺍﻟﱵ ﻭﻛﺒﺪﻳﻞ‪ ،‬ﺗﺘﻌﺎﻣﻞ ﻣﻊ ﺗﻠﻚ ﺍﳌﺘﻐﲑﺍﺕ ﺑﻨﻔﺲ‬
‫ﺍﻷﺳﻠﻮﺏ ﺩﻭﻥ ﺇﻗﺼﺎﺀ ﺃﻭ ﺍﻧﺘﻘﺎﺋﻴﺔ‪ .‬ﻳﻨﻌﻜﺲ ﻫﺬﺍ ﺍﻟﻔﻌﻞ ﰲ ﺇﺩﺭﺍﺝ ﻋﺪﺩ ﻛﺎﰲ‬
‫ﻭﻣﻨﺎﺳﺐ ﻣﻦ ﺍﳌﺘﻐﲑﺍﺕ ﺍﳌﺆﺧﺮﺓ ﻟﻜﻞ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﺪﺍﺧﻠﻴﺔ ﻭﰲ ﻛﻞ ﺍﳌﻌﺎﺩﻻﺕ‪.‬‬
‫ﺗﻌﺘﱪ ﳕﺎﺫﺝ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﺍﻟﺸﻌﺎﻋﻴﺔ ‪ VAR‬ﰲ ﺍﻷﺻﻞ ﳕﺎﺫﺝ ﺳﻼﺳﻞ‬
‫ﺯﻣﻨﻴﺔ ﻣﺘﻌﺪﺩﺓ "‪ "multivariate models‬ﺑﺴﻴﻄﺔ ﻻ ﻳﺘﻢ ﻓﻴﻬﺎ ﺗﻔﺴﲑ ﺍﳌﺘﻐﲑ ﺍﻟﺘﺎﺑﻊ‬
‫ﺳﻮﻯ ﺑﺴﻠﻮﻛﻪ ﺍﳌﺎﺿﻲ ﻭﻣﺎﺿﻲ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﺪﺍﺧﻠﻴﺔ ﺍﻷﺧﺮﻯ ﺍﳌﻘﺤﻤﺔ ﰲ ﺍﻟﻨﻤﻮﺫﺝ‪.‬‬
‫ﻟﺘﻮﺿﻴﺢ ﺍﻟﺮﺅﻳﺔ‪ ،‬ﻧﻔﺘﺮﺽ ﺃﻧﻪ ﻟﺪﻳﻨﺎ ﳕﻮﺫﺝ ﺫﻭ ﻣﺘﻐﲑﻳﻦ ﺩﺍﺧﻠﻴﲔ ﻓﻘﻂ ﻣﻊ‬
‫ﺩﺭﺟﱵ ﺗﺄﺧﲑ )‪ (lag of two periods‬ﺍﳌﺼﻮﺭﻳﻦ ﰲ ﺍﳌﺜﺎﻝ ﺍﳌﺨﺘﺼﺮ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪26 - Vector Auto Regression models (V.A.R).‬‬

‫‪203‬‬
 ‫‪y t = b 1 x t -1 + b 2 x t - 2 + b 3 y t -1 + b 4 y t - 2 + e 1 t‬‬
‫)‪(5.47‬‬
‫‪x t = a 1 x t -1 + a 2 x t - 2 + a 3 y t -1 + a 4 y t - 2 + e 2 t‬‬

‫ﻭ ‪ xt‬ﺑﻮﺍﺳﻄﺔ‬ ‫ﺣﻴﺚ ﺗﻔﺴﺮ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺘﻐﲑ ﺍﻟﺘﺎﺑﻊ ﻟﻜﻞ ﻣﻌﺎﺩﻟﺔ ﻫﻨﺎ‬
‫‪yt‬‬
‫ﺳﻠﻮﻛﻪ ﺍﳌﺎﺿﻲ ﺇﺿﺎﻓﺔ ﻟﻠﺴﻠﻮﻙ ﺍﳌﺎﺿﻲ ﻟﻠﻤﺘﻐﲑﺍﺕ ﺍﻷﺧﺮﻯ ‪(Lagged‬‬
‫)‪ Dependent Variables‬ﻣﻊ ﻣﺘﻐﲑﺍﺕ ﻣﺴﺘﻘﻠﺔ ﺇﻥ ﻭﺟﺪﺕ‪ .‬ﲢﺪﺩ ﺩﺭﺟﺔ ﺍﻟﺘﺄﺧﲑ‬
‫ﻋﻠﻰ ﺃﺳﺎﺱ ﺇﺣﺼﺎﺋﻲ ﻻ ﺩﺧﻞ ﻟﻠﻨﻈﺮﻳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻓﻴﻪ‪ .‬ﻭﺍﳌﻠﻔﺖ ﻟﻼﻧﺘﺒﺎﻩ ﰲ ﻫﺬﻩ‬
‫ﺍﳊﺎﻟﺔ‪ ،‬ﺃﻧﻪ ﳝﻜﻦ ﺗﻘﺪﻳﺮ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺑﻮﺍﺳﻄﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﺍﻟﻌﺎﺩﻳﺔ ﻟﻜﻞ‬
‫ﻣﻌﺎﺩﻟﺔ ﻋﻠﻰ ﺣﺪﺍ‪.‬‬
‫ﺣﻴﺚ ‪ e 1t‬ﻭ ‪ e 2t‬ﳝﺜﻼﻥ ﻣﺘﻐﲑﻳﻦ ﻋﺸﻮﺍﺋﻴﲔ ﻳﺘﺒﻌﺎﻥ ﺗﻮﺯﻳﻌﺎ ﻃﺒﻴﻌﻴﺎ ﻭﺗﺒﺎﻳﻦ‬
‫ﺛﺎﺑﺖ ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻦ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﳌﻌﺎﺩﻻﺕ ﻭ ﺑﺸﻜﻞ ﻋﺎﻡ ﻛﺎﻵﰐ‪:‬‬
‫‪yt = f (l ) yt + å t‬‬ ‫)‪(5.48‬‬

‫ﺃﻳﻦ ﺃﺑﻌﺎﺩ ﺍﻟﺸﻌﺎﻉ ‪ yt‬ﻭ ‪ St‬ﻫﻲ )‪ ( n ´ 1‬ﻟﻜﻞ ﻣﺘﻐﲑﺍﺕ ﺍﻟﻨﻤﻮﺫﺝ ﻣﻊ‬

‫ﺍﻋﺘﺒﺎﺭ ﺃﻥ ‪ St‬ﲤﺜﻞ ﺷﻌﺎﻉ ﻣﻘﺎﺩﻳﺮ ﺍﻷﺧﻄﺎﺀ‪ .‬ﲤﺜﻞ ) ‪ f (l‬ﻣﺼﻔﻮﻓﺔ ﻛﺜﲑ ﺍﳊﺪﻭﺩ ﰲ‬
‫ﻣﻌﺎﻣﻞ ﺍﻟﺘﺄﺧﲑ‪ L‬ﻣﻦ ﺍﻟﺪﺭﺟﺔ ‪ P‬ﺫﻭ ﺍﻷﺑﻌﺎﺩ )‪. (n ´ n‬‬
‫ﻳﺘﺒﻊ ﺗﻮﺯﻳﻌﺎﹰ ﻃﺒﻴﻌﻴﺎﹰ ﺑﻮﺳﻂ ﻣﻌﺪﻭﻡ ﻭﻣﺼﻔﻮﻓﺔ ﺗﺒﺎﻳﻨﺎﺗﻪ‬ ‫ﺃﻥ ‪St‬‬ ‫ﺑﺎﻓﺘﺮﺍﺽ‬
‫‪ W‬ﺃﻱ‪:‬‬
‫)‪S t ® N (0, W‬‬

‫ﳝﻜﻦ ﺇﻋﺎﺩﺓ ﻛﺘﺎﺑﺔ )‪ (2‬ﰲ ﺍﻟﺸﻜﻞ ﺍﳌﺨﺘﺼﺮ )‪ (reduced form‬ﺍﻟﺘﺎﱄ‪:‬‬

‫‪y t = F (l ) -1 å t‬‬ ‫)‪(5.49‬‬

‫ﻳﻌﺘﱪ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﺷﻜﻞ ﻣﻦ ﺃﺷﻜﺎﻝ ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‪.‬‬

‫‪204‬‬
‫ﳝﻜﻦ ﺗﻘﺪﻳﺮ ﺍﻟﻨﻤﻮﺫﺝ )‪ (5.47‬ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﻟﻜﻞ ﻣﻌﺎﺩﻟﺔ ﻋﻠﻰ‬
‫ﺣﺪﺍ‪ ،‬ﻧﻈﺮﺍ ﻟﻜﻮﻥ ﺍﻟﻄﺮﻑ ﺍﻷﳝﻦ ﻟﻠﻤﻌﺎﺩﻟﺘﲔ ﻣﺘﺸﺎﺑﻪ ﲤﺎﻣﺎ ﻻ ﳜﺘﻠﻒ ﺇﻻ ﻣﻦ ﺣﻴﺚ‬
‫ﺍﳋﻄﺄ ﺍﻟﻌﺸﻮﺍﺋﻲ‪.‬‬
‫ﺭﻏﻢ ﺃﳘﻴﺘﻬﺎ ﻭﺍﻻﻧﻔﺮﺍﺝ ﺍﻟﺬﻱ ﺃﺣﺪﺛﺘﻪ ﻗﻲ ﳎﺎﻝ ﺍﻟﻨﻤﺬﺟﺔ ﺍﻟﻘﻴﺎﺳﻴﺔ ﺇﻻ ﺃﻧﻪ‬
‫ﻳﻌﺎﺏ ﻋﻠﻴﻬﺎ ﻣﺎ ﻳﻠﻲ‪:‬‬

‫‪ -‬ﻛﺜﺮﺓ ﺍﳌﻌﻠﻤﺎﺕ ﺍﳌﻮﺿﻮﻋﺔ ﻟﻠﺘﻘﺪﻳﺮ ‪ :over parameterised‬ﺇﺿﺎﻓﺔ‬

‫ﻹﺷﻜﺎﻟﻴﺔ ﲢﺪﻳﺪ ﺩﺭﺟﺔ ﺍﻟﺘﺄﺧﲑ ﰲ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ‪ ،‬ﻳﺒﻘﻰ ﻋﺪﺩ ﺍﳌﻌﻠﻤﺎﺕ ﻣﻌﺘﱪﺍ ﺟﺪﺍ‬
‫ﺇﺫ ﻳﻌﺎﺩﻝ ) ‪ ( n 2 L + g‬ﺣﻴﺚ ‪ n, L , g‬ﲤﺜﻞ ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ ﻋﺪﺩ ﺍﳌﻌﺎﺩﻻﺕ‪ ،‬ﺩﺭﺟﺔ‬
‫ﺍﻟﺘﺄﺧﲑ ﻭﻋﺪﺩ ﺍﳌﺘﻐﲑﺍﺕ ﺍﳌﺴﺘﻘﻠﺔ‪ .‬ﻓﻘﺪ ﺑﻠﻎ ﻋﺪﺩ ﺍﳌﻌﻠﻤﺎﺕ ﰲ ﺍﻟﻨﻤﻮﺫﺝ)‪(5.47‬‬
‫ﺃﻋﻼﻩ)‪ (8‬ﺑﺪﻭﻥ ﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ‪ ،‬ﺑﻴﻨﻤﺎ ﻗﺪﺭ ﰲ ﳕﻮﺫﺝ )‪ (Sims1980‬ﻣﻊ ﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ‬
‫ﺏ ‪ 294‬ﻣﻌﻠﻤﺔ ﺃﻱ) ‪ ،( 6 2 ´ 8 + 6‬ﻭﻫﻮ ﺭﻗﻢ ﳝﺜﻞ ﻋﺪﺩﺍﹰ ﻫﺎﺋﻼﹰ ﻣﻦ ﺍﳌﻌﻠﻤﺎﺕ ﺍﻟﱵ‬
‫ﲢﺘﺎﺝ ﺇﱃ ﺗﻘﺪﻳﺮ ﻭﺗﻔﺴﲑ ﻭ ﺗﻘﻴﻴﻢ‪.‬‬
‫‪ -‬ﺍﻟﺘﻌﺪﺩ ﺍﳋﻄﻲ‪ :‬ﻳﻮﺟﺪ ﺍﺣﺘﻤﺎﻝ ﻛﺒﲑ ﻹﻣﻜﺎﻧﻴﺔ ﻣﻌﺎﻧﺎﺓ ﺍﳌﻌﺎﺩﻻﺕ ﻣﻦ‬
‫ﻣﺸﻜﻞ ﺍﻟﺘﻌﺪﺩ ﺍﳋﻄﻲ ﺑﲔ ﺍﳌﺘﻐﲑﺍﺕ ﺍﳌﺆﺧﺮﺓ‪ ،‬ﳑﺎ ﻳﻨﺘﺞ ﻋﻨﻪ ﻣﻘﺪﺭﺍﺕ ﻏﲑ ﺩﻗﻴﻘﺔ‬
‫)‪.(Imprecise‬‬
‫‪ -‬ﺿﻌﻒ ﺍﻟﺘﻨﺒﺆ‪ :‬ﺇﻥ ﻫﺬﺍ ﺍﻟﻌﺪﺩ ﺍﳍﺎﺋﻞ ﻣﻦ ﺍﳌﻌﻠﻤﺎﺕ ﺿﻤﻦ ﻛﻞ ﻣﻌﺎﺩﻟﺔ ﻗﺪ‬
‫ﻳﻌﻄﻲ ﻗﻮﺓ ﺇﺣﺼﺎﺋﻴﺔ ﺟﻴ‪‬ﺪﺓ ﻣﻘﺎﺑﻞ ﺗﻨﺒﺆ ﺿﻌﻴﻒ‪.‬‬
‫ﻟﺘﻔﺎﺩﻱ ﻫﺬﻩ ﺍﳌﺸﺎﻛﻞ‪ ،‬ﺍﺿﻄﺮ ﺳﻴﻤﺲ ﻓﻴﻤﺎ ﺑﻌﺪ ﺇﱃ ﺇﻋﺎﺩﺓ ﺍﻻﻋﺘﺒﺎﺭ ﻭﺇﻗﺤﺎﻡ‬
‫ﺍﻟﻨﻈﺮﻳﺔ ﺍﻻﻗﺘﺼﺎﺩﻳﺔ ﻟﺘﺨﻔﻴﺾ ﻋﺪﺩ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ‪ ،‬ﻛﻤﺎ ﺃﻥ ﺍﳊﻜﻢ ﻋﻠﻰ ﺃﳘﻴﺔ‬
‫ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﻳﺘﻄﻠﺐ ﺩﺭﺍﺳﺔ ﻭﻣﺘﺎﺑﻌﺔ ﻣﻌﻤﻘﺔ ﻟﺘﻄﺒﻴﻘﺎ‪‬ﺎ‪ ،‬ﰲ ﻫﺬﺍ ﺍﻟﺸﺄﻥ‪ ،‬ﻟﻘﺪ‬
‫ﻃﺮﺣﺖ ﻋﺪﺓ ﺃﺳﺌﻠﺔ ﳏﻮﺭﻳﺔ ﺃﳘﻬﺎ‪:‬‬
‫‪ -‬ﻫﻞ ﺃﻥ ﺗﻄﺒﻴﻖ ﺗﻘﻨﻴﺔ ‪ V . A.R.‬ﻳﻘﺘﻀﻲ ﺃﻭ ﻳﺸﺘﺮﻁ ﺍﺳﺘﺨﺪﺍﻡ ﻣﺘﻐﲑﺍﺕ‬
‫ﻣﺴﺘﻘﺮﺓ ﺃﻡ ﻻ؟‪ .‬ﻟﻘﺪ ﺍﺳﺘﺨﺪﻡ ﺳﻴﻤﺲ )‪(1980‬اﳌﺘﻐﲑﺍﺕ ﺍﻷﺻﻠﻴﺔ )‪ (levels‬ﺣﱴ‬

‫‪205‬‬
‫ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﺘﻐﲑﺍﺕ ﻏﲑ ﺍﳌﺴﺘﻘﺮﺓ ﻛﺎﻟﻜﺘﻠﺔ ﺍﻟﻨﻘﺪﻳﺔ ﻭﺍﻷﺳﻌﺎﺭ‪ ،‬ﺑﻴﻨﻤﺎ ﺍﺳﺘﺨﺪﻡ‬
‫ﺍﻵﺧﺮﻭﻥ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﺍﳌﺴﺘﻘﺮﺓ‪ .‬ﰲ ﺣﺎﻟﺔ ﺍﻟﺴﻼﺳﻞ ﻏﲑ ﺍﳌﺴﺘﻘﺮﺓ‪ ،‬ﻣﺎ ﻫﻲ‬
‫ﺍﻟﺘﻘﻨﻴﺔ ﺍﳌﻨﺎﺳﺒﺔ ﻹﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ ،‬ﺃﺑﻄﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ ﺃﻡ ﺑﻄﺮﻳﻘﺔ ﺍﻟﺘﻘﺪﻳﺮ‬
‫ﳌﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫‪ -‬ﻫﻞ ﺗﺘﻤﻴﺰ ﻫﺬﻩ ﺍﳌﺘﻐﲑﺍﺕ ﺑﺎﻟﺘﻜﺎﻣﻞ ﺍﳌﺸﺘﺮﻙ ‪Co-Integration‬؟ ﻭﻛﻴﻒ‬
‫ﳝﻜﻦ ﺍﺧﺘﺒﺎﺭ ﺫﻟﻚ ؟‬
‫ﻟﻘﺪ ﺍﺟﺘﻤﻊ ﺍﻟﻜﺜﲑ ﻋﻠﻰ ﺿﺮﻭﺭﺓ ﺗﻮﻓﺮ ﻣﺘﻐﲑﺍﺕ ﻣﺴﺘﻘﺮﺓ ﺣﱴ ﻳﺘﻢ ﳕﺬﺟﺘﻬﺎ‬
‫ﻭﻣﻦ ﰒ ﺍﻟﺘﻌﺎﻣﻞ ﻣﻌﻬﺎ ﰲ ﺷﻜﻠﻬﺎ ﺍﻷﺻﻠﻲ )‪ .(In levels‬ﺃﻣﺎ ﰲ ﺣﺎﻟﺔ‬
‫ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﻏﲑ ﺍﳌﺴﺘﻘﺮﺓ ﺍﻟﱵ ﺗﻜﻮﻥ ﻣﺼﺪﺭ ﻟﻈﺎﻫﺮﺓ ﺍﻟﺘﻘﺪﻳﺮ ﺍﳌﻠﻔﻖ‬
‫"‪ "spurious regression‬ﺍﻟﱵ ﺃﺷﺎﺭ ﺇﻟﻴﻬﺎ ﻛﻞ ﻣﻦ ‪ Newbold‬ﻭ ‪Granger‬‬
‫)‪ (1974‬ﺍﻟﱵ ﲢﺪﺙ ﺃﺛﻨﺎﺀ ﳏﺎﻭﻟﺔ ﺗﻘﺪﻳﺮ ﻋﻼﻗﺔ ﻭﳘﻴﺔ ﺧﺎﺻﺔ ﺑﺎﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﻏﲑ‬
‫ﺍﳌﺴﺘﻘﺮﺓ ﺃﻭ ﺍﻟﱵ ﻻ ﻋﻼﻗﺔ ﻣﻨﺘﻈﺮﺓ ﺃﻭ ﻣﺘﻮﻗﻌﺔ ﺑﲔ ﺍﳌﺘﻐﲑﻳﻦ ﺍﻟﺘﺎﺑﻊ ﻭ ﺍﳌﺴﺘﻘﻞ ﺃﻱ‬
‫ﻋﻠﻰ ﺳﺒﻴﻞ ﺍﳌﺜﺎﻝ ﻻ ﺍﳊﺼﺮ‬
‫‪y t = b 1 + b 2 xt + e t‬‬ ‫)‪(5.50‬‬

‫ﺣﻴﺚ ‪ yt‬ﻭ ‪ xt‬ﺳﻠﺴﻠﺘﲔ ﻋﺸﻮﺍﺋﻴﺘﲔ ﻏﲑ ﻣﺴﺘﻘﺮﺗﲔ ﺑﺴﺒﺐ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬

‫ﻓﻴﻜﻮﻥ ﺍﻟﺘﻘﺪﻳﺮ ﻭﳘﻴﺎ ﻛﻮﻧﻪ ﻳﻌﻄﻲ ﻧﺘﺎﺋﺞ ﻇﺎﻫﺮﻫﺎ ﺟﻴﺪ ﻣﻦ ﺣﻴﺚ ﺟﻮﺩﺓ ﺍﻟﺘﻮﻓﻴﻖ‬
‫ﻭﻣﻌﻨﻮﻳﺔ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺇﺣﺼﺎﺋﻴﱵ )‪ (t‬ﻭ)‪ (F‬ﻭ ﻟﻜﻦ ﺑﺎﻃﻨﻬﺎ ﻭﳘﻲ‬
‫ﻻ ﳛﻤﻞ ﻣﻌﲎ ﺃﻭ ﻣﻐﺰﻯ ﺍﻗﺘﺼﺎﺩﻱ‪ ،‬ﻛﻮﻥ ‪ xt‬ﰲ ﻫﺬﻩ ﺍﻟﻈﺮﻭﻑ ﺗﻜﻮﻥ ﺳﲑﻭﺭﺓ ﻏﲑ‬
‫ﻣﺴﺘﻘﺮﺓ ﻭﺗﺼﺒﺢ ﻓﻴﻪ ﺍﳌﺼﻔﻮﻓﺔ ) ‪ ( X ¢X‬ﻏﲑ ﺁﻳﻠﺔ ﳓﻮ ﺃﻱ ﺗﻮﺯﻳﻊ ‪‬ﺎﺋﻲ ﻭ ﻣﻨﻪ ﻳﻜﻮﻥ‬
‫‪T‬‬
‫ﻣﻘﺪﺭ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﻏﲑ ﻣﺘﻨﺎﺳﻖ "‪" not consistent‬ﻛﻤﺎ ﺗﺼﺒﺢ ﻃﺮﻕ‬
‫ﺍﻻﺳﺘﺪﻻﻝ ﺍﳌﺄﻟﻮﻓﺔ ﻏﲑ ﺻﺤﻴﺤﺔ‪ .‬ﳍﺬﺍ ﺍﻗﺘﺮﺡ ﻛﻞ ﻣﻦ ‪ Granger‬ﻭ ‪ Newbold‬ﺃﻥ‬
‫ﻳﺘﻢ ﺍﻟﺘﻘﺪﻳﺮ ﺑﻔﺮﻭﻕ ﺍﳌﺘﻐﲑﺍﺕ ﺑﺪﻝ ﺍﳌﺴﺘﻮﻳﺎﺕ ﻋﻨﺪﻣﺎ ﻳﺆﺩﻱ ﺗﻘﺪﻳﺮ ﺍﳌﺮﺑﻌﺎﺕ‬
‫ﺍﻟﺼﻐﺮﻯ ﺇﱃ ‪ R‬ﻛﺒﲑ ﻭ ‪ DW‬ﺿﻌﻴﻒ‪ ،‬ﺃﻱ‪:‬‬‫‪2‬‬

‫‪206‬‬
 ‫‪Dyt = b1 + b 2 Dxt + vt‬‬ ‫)‪(5.51‬‬

‫ﺃﻳﻦ ‪ ، vt = e t - e t -1‬ﻭﺗﺼﺒﺢ ﻣﺮﺗﺒﻄﺔ ﺗﺴﻠﺴﻠﻴﺎ ﻭﻟﻜﻦ ﻣﻘﺪﺭ ﺍﳌﺮﺑﻌﺎﺕ‬

‫ﺍﻟﺼﻐﺮﻯ ﻝ ‪ b 2‬ﻳﺒﻘﻰ ﻣﻨﺴﺠﻤﺎﹰ ﺭﻏﻢ ﲢﻴﺰﻩ ﻭﻋﺪﻡ ﻓﻌﺎﻟﻴﺘﻪ‪ .‬ﰲ ﻫﺬﻩ ﺍﻟﻮﺿﻌﻴﺔ ﻳﻜﻮﻥ‬
‫ﺗﻘﺪﻳﺮ ﺗﻠﻚ ﺍﳌﻌﻠﻤﺔ ﺃﺣﻮﻁ ﺑﺎﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ ﻣﻦ ﺍﳌﻌﺎﺩﻟﺔ )‪ (5.51‬ﺑﺪﻝ )‪.(5.50‬‬

‫ﺃﻣﺎ ﻣﺎ ﺗﻌﻠﻖ ﲟﻮﺿﻮﻉ ﲢﻮﻳﻞ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﻏﲑ ﺍﳌﺴﺘﻘﺮﺓ ﺇﱃ ﺃﺧﺮﻯ‬

‫ﺳﺎﻛﻨﺔ‪ ،‬ﳚﺐ ﻣﺒﺪﺋﻴﺎﹰ ﺍﻟﻔﺼﻞ ﰲ ﻃﺮﻳﻘﺔ ﺇﺯﺍﻟﺔ ﺳﺒﺐ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﺍﳌﺘﻤﺜﻞ‬
‫ﺑﺎﻷﺳﺎﺱ ﰲ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪ .‬ﻳﻌﱪ ﺭﻳﺎﺿﻴﺎ ﻋﻦ ﺍﻟﺴﻠﺴﻠﺔ ﻏﲑ ﺍﳌﺴﺘﻘﺮﺓ ﻭﺣﺴﺐ‬
‫ﺳﺒﺐ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﺍﳌﺘﻤﺜﻞ ﰲ ﳕﻮﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺃﻭ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ ﻭﻛﻤﺎ‬
‫ﻳﻠﻲ‪:‬‬
‫ﳕﻮﺫﺝ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪:‬‬
‫‪yt = f (t ) + e t‬‬

‫ﻭﺍﻟﺬﻱ ﻳﺄﺧﺬ ﺷﻜﻞ ﺍﻟﻌﻼﻗﺔ )‪ (3.1‬ﺇﺫﺍ ﻛﺎﻧﺖ ﺍﻟﺪﺍﻟﺔ ﺧﻄﻴﺔ‪:‬‬

‫‪yt = a + bt + e t‬‬

‫ﳕﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ ﺑﺪﻭﻥ ﺣﺪ ﺛﺎﺑﺖ )‪:(5.1‬‬

‫‪yt = yt -1 + e t‬‬

‫ﻭﺑﺴﺒﺐ ﺍﻟﺴﻠﻮﻙ ﺍﳌﺘﻤﺎﺛﻞ ﻭﺍﳌﺘﺸﺎﺑﻪ ﳍﺬﻳﻦ ﺍﻟﻨﻤﻮﺫﺟﲔ‪ ،‬ﳝﻜﻦ ﺍﻟﺘﺴﺎﺅﻝ ﻋﻦ‬

‫ﺍﻟﻄﺮﻳﻘﺔ ﺍﻷﻧﺴﺐ ﻹﺯﺍﻟﺔ ﺳﺒﺐ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﺍﳌﺘﻤﺜﻞ ﺑﺎﻷﺳﺎﺱ ﰲ ﺃﺣﺪ‬
‫ﺍﻟﻨﻤﻮﺫﺟﲔ ﺍﻟﺴﺎﺑﻘﲔ‪.‬‬

‫‪207‬‬
 ‫ﺍﺧﺘﺒﺎﺭ ﺍﳉﺬﺭ ﺍﻷﺣﺎﺩﻱ‬
‫ﺟﺎﺀ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻟﺘﺪﻋﻴﻢ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻜﻠﻴﺔ ﰲ ﲢﺪﻳﺪ ﻃﺒﻴﻌﺔ ﻋﺪﻡ‬
‫ﺍﺳﺘﻘﺮﺍﺭ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻭﳏﺎﻭﻟﺔ ﺍﺧﺘﻴﺎﺭ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻷﻧﺴﺐ ﻹﺯﺍﻟﺔ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ‬
‫ﺳﻮﺍﺀ ﻋﻦ ﻃﺮﻳﻖ ﺇﺑﻌﺎﺩ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ )‪ (detrending‬ﺃﻭ ﻋﻦ ﻃﺮﻳﻖ ﺍﻟﻔﺮﻭﻕ‬
‫)‪ .(differencing‬ﺗﺴﻤﺢ ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ ﺍﻷﺧﲑﺓ ﺑﺘﻔﺎﺩﻱ ﻣﺸﻜﻞ ﺍﻟﺘﻘﺪﻳﺮ ﺍﳌﺰﻳﻒ‬
‫ﻭﻣﻨﻪ ﺍﶈﺎﻓﻈﺔ ﻋﻠﻰ ﺧﺼﺎﺋﺺ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﻹﺣﺼﺎﺋﻴﺔ ﲝﻜﻢ ﺃﻥ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﻏﲑ ﺍﳌﺴﺘﻘﺮﺓ ﺗﺸﻮﻩ ﺧﺼﺎﺋﺺ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﻟﺘﻘﻠﻴﺪﻳﺔ ﺍﳌﻌﺮﻭﻓﺔ ﻛﻤﺎ ﺗﻔﻘﺪ ﺗﻮﺯﻳﻌﺎ‪‬ﺎ‬
‫ﺍﳌﻌﺘﺎﺩﺓ‪.‬‬
‫)‪(Time trend‬‬ ‫ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‬ ‫‪.1‬‬

‫ﺑﺎﻟﺮﺟﻮﻉ ﺇﱃ ﺍﻟﻨﻤﻮﺫﺝ )‪ (3.1‬ﺍﳌﻌﺎﺩ ﻛﺘﺎﺑﺘﻪ ﺃﻋﻼﻩ‪ ،‬ﻭ ﺑﻌﺪ ﺗﻘﺪﻳﺮﻩ ﰲ ﺍﻟﺸﻜﻞ‬

‫ﺍﳋﻄﻲ ﻳﺘﻢ ﺣﺴﺎﺏ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﳌﺮﻛﺒﺔ ﺍﻟﻔﺼﻠﻴﺔ ﻋﻦ ﻃﺮﻳﻖ ﺣﺴﺎﺏ ﺑﻮﺍﻗﻲ‬
‫ﺍﻟﺘﻘﺪﻳﺮ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪et = yt - aˆ + bˆt‬‬

‫ﺗﺼﺒﺢ ﺳﻠﺴﻠﺔ ﺍﻟﺒﻮﺍﻗﻲ ﺳﻠﺴﻠﺔ ﺧﺎﻟﻴﺔ ﻣﻦ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭ ﺑﺎﻟﺘﺎﱄ‬

‫ﺳﻠﺴﻠﺔ ﻣﺴﺘﻘﺮﺓ‪.‬‬

‫)‪(Random walk‬‬ ‫ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ‬ ‫‪.2‬‬

‫ﳝﻜﻦ ﺻﻴﺎﻏﺘﻪ ﺭﻳﺎﺿﻴﺎﹰ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‬

‫‪yt = b 0 + yt -1 + e t‬‬

‫ﻭﺑﺘﻄﺒﻴﻖ ﺗﻘﻨﻴﺔ ﺍﻟﻔﺮﻭﻕ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻳﺘﻢ ﺍﳊﺼﻮﻝ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﺴﺘﻘﺮﺓ‬
‫ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪208‬‬
 ‫‪Dyt = yt - yt -1 = b 0 + e t‬‬

‫ﺃﻱ ﺃﻥ ﻫﺬﻩ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳉﺪﻳﺪﺓ ﺗﺘﺬﺑﺬﺏ ﺣﻮﻝ ﺍﻟﻮﺳﻂ ﺍﳊﺴﺎﰊ ﺍﻟﺜﺎﺑﺖ ‪. b 0‬‬
‫ﻋﻠﻰ‬ ‫‪t=1,T‬‬ ‫ﻟﺘﺒﻴﺎﻥ ﲤﺎﺛﻞ ﺳﻠﻮﻙ ﺍﻟﻨﻤﻮﺫﺟﲔ‪ ،‬ﳒﺮﻱ ﺍﻟﺘﺤﻮﻳﻼﺕ ﺍﻟﺘﺎﻟﻴﺔ ﳌﺎ‬
‫ﳕﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ ﻭ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪y1 = b 0 + y0 + e 1‬‬

‫‪y2 = b 0 + y1 + e 2‬‬

‫) ‪= 2 b 0 + y0 + (e 1 + e 2‬‬

‫) ‪y3 = 3b 0 + y0 + (e 1 + e 2 + e 3‬‬

‫) ‪y4 = 4 b 0 + y0 + (e 1 + e 2 + e 3 + e 4‬‬

‫ﻭﻋﻨﺪ ﺍﻟﻔﺘﺮﺓ ‪ T‬ﻳﻜﺘﺐ ﺍﻟﻨﻤﻮﺫﺝ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪T‬‬
‫‪y t = y 0 + tb 0 + å e t‬‬
‫‪t =1‬‬

‫ﺃﻱ‪:‬‬
‫‪yt = y0 + b 0 (t ) + vt‬‬ ‫)‪(5.52‬‬

‫ﺍﳌﺮﺗﺒﻂ‬ ‫‪vt‬‬ ‫ﺍﻟﱵ ﺗﺸﺒﻪ ﲤﺎﻣﺎﹰ ﻣﻌﺎﺩﻟﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻓﻴﻤﺎ ﻋﺪﺍ ﺍﳋﻄﺄ ﺍﻟﻌﺸﻮﺍﺋﻲ‬
‫ﺫﻭ ﺗﺒﺎﻳﻦ ﻳﻌﺎﺩﻝ ‪ ts 2‬ﺍﻟﺬﻱ ﻳﺴﺘﻠﺰﻡ ﺃﻳﻀﺎﹰ ﻋﺪﻡ ﺍﻻﺳﺘﻘﺮﺍﺭ ﰲ ﺍﻟﺘﺒﺎﻳﻦ‪.‬‬
‫)‪(5.50‬‬ ‫ﺇﺫﺍ ﺃﺭﺩﻧﺎ ﺧﻄﺄﹰ ﺇﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺑﻌﺪ ﺗﻘﺪﻳﺮﻫﺎ ﻣﻦ ﺍﳌﻌﺎﺩﻟﺔ‬
‫ﻭﻓﻖ ﺍﳌﻌﺎﺩﻟﺔ )‪ (3.1‬ﺃﻭ )‪ (5.1‬ﻧﻜﻮﻥ ﻗﺪ ﻭﻗﻌﻨﺎ ﰲ ﻣﺸﻜﻠﺔ ﺍﻟﺘﻘﺪﻳﺮ ﺍﻟﻮﳘﻲ‬
‫)‪ (Spurious regression‬ﻷﻥ ﻛﻠﺘﺎ ﺍﻟﺴﻠﺴﻠﺘﲔ ‪ yt‬ﻭ ) ‪ (t‬ﻏﲑ ﻣﺴﺘﻘﺮﺗﲔ‪ ،‬ﻭﻣﻨﻪ‬
‫ﺗﺘﺤﻘﻖ ﺍﻟﺴﻠﺒﻴﺎﺕ ﺍﳌﺬﻛﻮﺭﺓ ﺳﺎﺑﻘﺎﹰ‪.‬‬

‫‪209‬‬
‫)‪(Dickey and Fuller‬‬ ‫ﻟﺘﻔﺎﺩﻱ ﻫﺬﻩ ﺍﻟﻮﺿﻌﻴﺎﺕ‪ ،‬ﻳﺴﺘﺨﺪﻡ ﺍﺧﺘﺒﺎﺭ‬
‫ﻟﺘﺤﺪﻳﺪ ﻓﻴﻤﺎ ﺇﺫﺍ ﻛﺎﻧﺖ ﺍﻟﺴﻠﺴﻠﺔ ﻗﺎﺑﻠﺔ ﻟﺘﺼﺒﺢ ﰲ ﺷﻜﻞ ﻣﺴﺘﻘﺮ ﻋﻦ ﻃﺮﻳﻖ ﺗﻘﻨﻴﺔ‬
‫ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺃﻭ ﺍﻟﻔﺮﻭﻕ ﺍﳌﺬﻛﻮﺭﺗﲔ ﺳﺎﺑﻘﺎ‪.‬‬
‫ﻳﺘﻮﻗﻒ ﻫﺬﺍ ﺍﻻﺧﺘﺒﺎﺭ ﻋﻠﻰ ﺍﻟﺼﻴﺎﻏﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪yt = a + bt + ryt -1 + e t‬‬ ‫)‪(5.53‬‬

‫ﻓﺈﺫﺍ ﺛﺒﺖ ﻋﻦ ﻃﺮﻳﻖ ﺍﻻﺧﺘﺒﺎﺭ ﺃﻥ ‪ r = 1‬ﻭ ‪ b = 0‬ﺗﺼﺒﺢ ﺍﻟﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮﺓ‬

‫ﻋﻦ ﻃﺮﻳﻖ ﺗﻘﻨﻴﺔ ﺍﻟﻔﺮﻭﻕ )‪ ، (Difference stationary‬ﺃﻣﺎ ﺇﺫﺍ ﻛﺎﻥ ‪ / r / < 1‬ﻓﺈﻥ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ‪ yt‬ﺗﺘﺤﻮﻝ ﺇﱃ ﻣﺴﺘﻘﺮﺓ ﻋﻦ ﻃﺮﻳﻖ ﺇﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ‪(Trend‬‬
‫)‪ stationary‬ﻛﻤﺎ ﺭﺃﻳﻨﺎ ﺿﻤﻦ ﻓﺼﻮﻝ ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ‪.‬‬
‫ﺇﻥ ﺗﻘﺪﻳﺮ ﺍﻟﻌﻼﻗﺔ ﺃﻋﻼﻩ )‪ (5.53‬ﺗﻌﲏ ﺇﺟﺮﺍﺀ ﺿﻤﲏ ﻟﺘﻘﺪﻳﺮ ﻣﺰﻳﻒ ﻛﻮﻥ‬
‫ﺍﻟﺴﻠﺴﻠﺘﲔ ‪ yt‬ﻭ ‪ t‬ﻏﲑ ﻣﺴﺘﻘﺮﺗﲔ‪ .‬ﳍﺬﺍ ﻓﻤﻦ ﺍﻷﻧﺴﺐ ﺍﻋﺘﻤﺎﺩ ﻃﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ ﻋﻨﺪ‬
‫ﺇﺯﺍﻟﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‪.‬‬
‫‪yt -1‬‬ ‫ﻟﺘﻔﺎﺩﻱ ﻫﺬﻩ ﺍﻟﻮﺿﻌﻴﺔ‪ ،‬ﳝﻜﻦ ﻃﺮﺡ ﻣﻦ ﻃﺮﰲ ﺍﳌﻌﺎﺩﻟﺔ ﺍﻟﺴﺎﺑﻘﺔ ﺍﳌﻘﺪﺍﺭ‬
‫ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪yt - yt -1 = a + b.t + ryt -1 - yt -1 + e t‬‬
‫ﻣﻨﻪ‪:‬‬
‫‪Dyt = a + b.t + ( r - 1) yt -1 + e t‬‬ ‫)‪(5.54‬‬

‫ﻛﻤﺎ ﻧﺴﺘﻄﻴﻊ ﺑﻌﺪ ﺍﻻﺳﺘﻌﺎﻧﺔ ﺑﺎﺧﺘﺒﺎﺭﻱ ‪ AIC‬ﻭ ‪ BIC‬ﲢﺪﻳﺪ )‪ (n‬ﻣﻘﺎﺩﻳﺮ‬

‫ﺇﺿﺎﻓﻴﺔ)‪ (lagged differences‬ﺑﺪﺭﺟﺎﺕ ﺃﻋﻠﻰ ﻣﻦ ﻃﺮﻳﻘﺔ ﺍﻻﳓﺪﺍﺭ ﺍﻟﺬﺍﰐ ﻗﺼﺪ‬
‫ﺍﻣﺘﺼﺎﺹ ﺗﺄﺛﲑ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﶈﺘﻤﻞ ﺑﲔ ﺍﻷﺧﻄﺎﺀ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪210‬‬
 ‫‪n‬‬
‫‪Dy t = a + b.t + ( r - 1) y t -1 + å r j D y t - j + e t‬‬ ‫)‪(5.55‬‬
‫‪j‬‬

‫‪j =1‬‬ ‫ﻓﻠﻤﺎ‬

‫‪Dyt = a + b.t + ( r - 1) yt -1 + r 1 ( yt -1 - yt - 2 ) + e t‬‬ ‫)‪(5.56‬‬

‫ﻳﻮﻓﺮ‬ ‫)‪(5.55‬‬
‫ﺃﻭ ﺍﳌﻌﺎﺩﻟﺔ‬ ‫ﺇﻥ ﺗﺸﻜﻴﻞ ﺍﻻﺧﺘﺒﺎﺭ ﻋﻠﻰ ﺃﺳﺎﺱ ﺍﳌﻌﺎﺩﻟﺔ‬
‫)‪(5.54‬‬

‫‪Augmented‬‬ ‫& ‪Dickey‬‬ ‫ﺍﺧﺘﺒﺎﺭﻱ )‪ Dickey & Fuller (DF‬ﺃﻭ‬

‫)‪ Fuller(ADF‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ ﻭﺣﺴﺐ ﺍﻟﺼﻴﺎﻏﺔ ﺍﳌﺸﺘﺮﻛﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪H 0 : r =1 , b = 0‬‬
‫‪HA : r á1‬‬ ‫‪,‬‬ ‫‪b¹0‬‬

‫ﳝﻜﻦ ﺗﺸﻜﻴﻞ ﺍﻻﺧﺘﺒﺎﺭ ﻭﻓﻖ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ ﻭ ﺍﻟﱵ ﺗﺄﺧﺬ ﻧﻔﺲ ﺷﻜﻞ ﺍﻟﻌﻼﻗﺔ‬
‫)‪:(3.26‬‬
‫‪( RRSS - URSS ) / m‬‬
‫=‪Q‬‬ ‫)‪(5.57‬‬
‫‪URSS / T - 3 - n‬‬

‫ﺣﻴﺚ ‪ RRSS‬و ‪ URSS‬ﲤﺜﻞ ﳎﻤﻮﻉ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ ﺍﳌﻘﻴﺪﺓ ﲢﺖ‬

‫ﺍﻟﻔﺮﺿﻴﺔ ‪ H 0‬ﻭﻏﲑ ﺍﳌﻘﻴﺪﺓ ﲢﺖ ‪ H A‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ ﺑﻴﻨﻤﺎ )‪ (m‬ﺗﻌﱪ ﻋﻦ ﻋﺪﺩ‬
‫ﺍﻟﻘﻴﻮﺩ ﺍﳌﻔﺮﻭﺿﺔ ﻭ ) ‪ (T - k‬ﺩﺭﺟﺎﺕ ﺣﺮ‪‬ﻳﺔ ﺃﻳﻦ )‪ (k=3‬ﲤﺜﻞ ﻋﺪﺩ ﻣﻌﻠﻤﺎﺕ‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻘﺪﺭ‪ .‬ﻳﻜﻤﻦ ﺍﻟﻔﺮﻕ ﺑﲔ ﺍﻟﻌﻼﻗﺘﲔ ‪ F‬ﻭ‪ Φ‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ ﰲ ﺍﺧﺘﻼﻑ‬
‫ﺗﻮﺯﻳﻌﻬﻤﺎ ﻓﻘﻂ‪.‬‬
‫ﺇﺫﺍ ﻛﺎﻧﺖ ‪ Φ‬ﺍﶈﺴﻮﺑﺔ ﺃﻗﻞ ﻣﻦ ﺗﻠﻚ ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ‬ ‫‪H0‬‬ ‫ﻳﺘﻢ ﻗﺒﻮﻝ ﺍﻟﻔﺮﺿﻴﺔ‬
‫ﺍﻟﻮﺍﺭﺩﺓ ﰲ ﺍﳉﺪﻭﻝ )‪ (37‬ﺃﺩﻧﺎﻩ‪.‬‬
‫‪Φ –statistic‬‬‫ﻳﺘﻤﺜﻞ ﺩﻭﺭ ﺍﺧﺘﺒﺎﺭ ‪ Dickey & Fuller‬ﰲ ﲢﺪﻳﺪ ﺗﻮﺯﻳﻊ‬
‫ﻟﻠﻔﺮﺿﻴﺔ ﺍﳌﺸﺘﺮﻛﺔ ﺍﳌﺬﻛﻮﺭﺓ ﺃﻋﻼﻩ‪ ،‬ﲝﻜﻢ ﺃﻥ ‪ yt‬ﺗﻜﻮﻥ ﻏﲑ ﻣﺴﺘﻘﺮﺓ ﻋﻨﺪﻣﺎ‬
‫ﻳﻜﻮﻥ ‪. r =1‬ﻭﻣﻨﻪ ﻓﺘﻘﺪﻳﺮ ﺍﻟﻌﻼﻗﺔ )‪ (5.53‬ﻳﻜﻮﻥ ﻣﻐﻠﻄﺎﹰ ﻭﻣﻨﻪ ﺗﺼﺒﺢ ﻛﻞ‬

‫‪211‬‬
‫‪Dickey & Fuller‬‬ ‫ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﻟﺘﻘﻠﻴﺪﻳﺔ ﺍﳌﺴﺘﺨﺪﻣﺔ ﻏﲑ ﻣﻘﺒﻮﻟﺔ‪ ،‬ﳍﺬﺍ ﺍﻗﺘﺮﺡ‬
‫ﺟﺪﻭﻝ ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ ﺍﻟﺘﺎﱄ‪:‬‬
‫ﺟﺪﻭﻝ ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ ﻟﺪﻳﻜﻲ ﻓﻮﻟﺮ‬ ‫ﺍﳉﺪﻭﻝ )‪:(37‬‬
‫ﺍﻟﻘﻴﻤﺔ ﺍﻟﺤﺭﺠﺔ ﻟـ ‪F‬‬ ‫ﺍﻟﻘﻴﻤﺔ ﺍﻟﺤﺭﺠﺔ ﻟـ ‪Φ‬‬ ‫ﺍﻟﻌﻴﻨﺔ‬

‫‪3.38‬‬ ‫‪7.24‬‬ ‫‪25‬‬

‫‪3.18‬‬ ‫‪6.73‬‬ ‫‪50‬‬
‫‪3.09‬‬ ‫‪6.49‬‬ ‫‪100‬‬
‫‪2.99‬‬ ‫‪6.25‬‬ ‫∞‬

‫ﻟﺘﻮﺿﻴﺢ ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ‪ ،‬ﻗﺎﻡ ﻛﻞ ﻣﻦ ‪ (1981) Dickey & Fuller‬ﺑﺪﺭﺍﺳﺔ‬

‫ﻟﻮﻏﺎﺭﻳﺘﻢ ﻣﺆﺷﺮ ﺍﻹﻧﺘﺎﺝ ﺍﻟﺼﻨﺎﻋﻲ ﺍﳌﻌﻄﻲ ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪ ،1977.4 -1950.1‬ﺣﻴﺚ‬
‫ﺍﻓﺘﺮﺿﺎ ﺃﻥ ﺍﻟﻨﻤﻮﺫﺝ )‪ (5.56‬ﻫﻮ ﺍﳌﻨﺎﺳﺐ ﻟﺘﻤﺜﻴﻞ ﻫﺬﻩ ﺍﳌﻌﻄﻴﺎﺕ‪:‬‬
‫‪Dyt = a + b.t + ( r - 1) yt -1 + r 1 ( yt -1 - yt - 2 ) + e t‬‬

‫ﺗﻜﻮﻳﻦ ﺍﻻﺧﺘﺒﺎﺭ‬
‫‪ (1‬ﻋﻠﻰ ﺃﺳﺎﺱ ﺍﻟﻔﺮﺿﻴﺔ ‪ HA‬ﺍﻟﺴﺎﺑﻘﺔ ﻳﺘﻢ ﺗﻘﺪﻳﺮ ﺍﻟﻨﻤﻮﺫﺝ ﺃﻋﻼﻩ‪:‬‬
‫) ‪Dyˆt = 0.52 + 0.0012.t - 0.119 yt -1 + 0.498( yt -1 - yt - 2‬‬
‫)‪(0.15) (0.00034) (0.033‬‬ ‫)‪(0.081‬‬
‫‪URSS = 0.056448‬‬
‫ﲢﺖ ﺍﻟﻔﺮﺿﻴﺔ ‪ H 0‬ﻛﻤﺎ ﻳﻠﻲ‪:‬‬ ‫‪ (2‬ﺇﻋﺎﺩﺓ ﺗﻘﺪﻳﺮ ﻧﻔﺲ ﺍﻟﻨﻤﻮﺫﺝ‬

‫) ‪Dyˆ t = 0.0054 + 0.447( yt -1 - yt - 2‬‬

‫)‪(0.0025) (0.083‬‬
‫‪RRSS = 0.063211‬‬

‫‪ (3‬ﺣﺴﺎﺏ ‪:Φ‬‬
‫‪( 0 . 063211 - 0 . 056448 ) / 2‬‬
‫= ‪Q‬‬ ‫‪= 6.34‬‬
‫‪0 . 056448 / 112 - 4‬‬

‫‪212‬‬
‫ﺑﺎﻟﺮﺟﻮﻉ ﺇﱃ ﺟﺪﻭﻝ ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ ﺍﳌﺬﻛﻮﺭ ﺃﻋﻼﻩ‪ ،‬ﳒﺪ ﺃﻥ ‪ Φ‬ﺍﶈﺴﻮﺑﺔ‬
‫ﺃﺻﻐﺮ ﻣﻦ ﺗﻠﻚ ﺍ‪‬ﺪﻭﻟﺔ )‪ (6.34 < 6.49‬ﻭﻣﻨﻪ ﻗﺒﻮﻝ ﺍﻟﻔﺮﺿﻴﺔ ‪ . H 0‬ﺑﺎﻟﺘﺎﱄ ﺗﻜﻮﻥ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻏﲑ ﻣﺴﺘﻘﺮﺓ ﻭ ﲢﻮﻳﻠﻬﺎ ﺇﱃ ﻣﺴﺘﻘﺮﺓ ﻳﺘﻢ ﻋﻦ ﻃﺮﻳﻖ ﺗﻘﻨﻴﺔ ﺍﻟﻔﺮﻭﻕ‬
‫ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‪.‬‬
‫ﻛﻤﺎ ﳝﻜﻦ ﺇﺟﺮﺍﺀ ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ ﻋﻦ ﻃﺮﻳﻖ ﺍﺧﺘﺒﺎﺭ ﺍﳌﻌﻠﻤﺔ ‪ r‬ﺍﻧﻄﻼﻗﺎﹰ ﻣﻦ‬
‫ﺍﻟﻌﻼﻗﺔ )‪ (5.54‬ﺍﻟﱵ ﻳﻌﺎﺩ ﺻﻴﺎﻏﺘﻬﺎ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪Dyt = a + b.t + dyt -1 + e t‬‬ ‫)‪(5.57‬‬

‫ﺣﻴﺚ )‪. d = (r - 1‬‬

‫ﻋﻠﻰ ﺇﺟﺮﺍﺀ ﺍﻻﳓﺪﺍﺭﺍﺕ ﺍﻟﺘﺎﻟﻴﺔ ﻭﻛﻤﺎ‬ ‫‪Dickey‬‬ ‫ﻳﺘﻮﻗﻒ ﺍﺧﺘﺒﺎﺭ ‪& Fuller‬‬
‫ﻳﻠﻲ‪:‬‬
‫‪ /1‬ﳕﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ ﺑﺪﻭﻥ ﺣﺪ ﺛﺎﺑﺖ‪:‬‬
‫‪Dyt = dyt -1 + e t‬‬
‫‪ /2‬ﳕﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ ﲝﺪ ﺛﺎﺑﺖ‪:‬‬
‫‪Dyt = a + dyt -1 + e t‬‬
‫‪ /3‬ﳕﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ ﲝﺪ ﺛﺎﺑﺖ ﻭﺍﲡﺎﻩ ﻋﺎﻡ‪:‬‬
‫‪Dyt = a + b.t + dyt -1 + e t‬‬

‫ﻳﺮﺟﻊ ﻫﺬﺍ ﺍﻟﺘﻔﺼﻴﻞ ﰲ ﺍﻟﻨﻤﺎﺫﺝ ﺇﱃ ﺇﺩﺭﺍﻙ ﺃﻥ ﺗﻮﺯﻳﻊ ﺩﻳﻜﻲ ﻓﻮﻟﺮ ﻳﺘﺄﺛﺮ‬

‫ﺑﺈﺿﺎﻓﺔ ﺍﳌﻘﺎﺩﻳﺮ ﺍﻟﺜﺎﺑﺘﺔ ﻭﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﻛﺬﺍ ﺍﳌﺘﻐﲑﺍﺕ ﺍﻟﺘﻤﺜﻴﻠﻴﺔ‪ .‬ﻋﻠﻰ ﻫﺬﺍ ﺍﻷﺳﺎﺱ‬
‫ﺗﺼﺎﻍ ﻓﺮﺿﻴﺔ ﺍﻻﺧﺘﺒﺎﺭ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫)ﻭﺟﻮﺩ ﺍﳉﺬﺭ ﺍﻷﺣﺎﺩﻱ( ‪H 0 : d = 0‬‬
‫)ﺳﻠﺴﻠﺔ ﻣﺴﺘﻘﺮﺓ( ‪HA : d ¹ 0‬‬

‫‪213‬‬
‫ﻳﻔﻀﻞ ﺇﺟﺮﺍﺀ ﺍﻻﺧﺘﺒﺎﺭ ﺣﺴﺐ ﺍﻟﺘﺮﺗﻴﺐ ﺃﻋﻼﻩ ﺑﺎﻻﺳﺘﻌﺎﻧﺔ ﺑﺈﺣﺼﺎﺀﺓ‬
‫ﺳﺘﻴﻮﺩﻧﺖ ﺍﻟﺘﻘﻠﻴﺪﻳﺔ ﺍﻟﱵ ﺃﺻﺒﺤﺖ ﰲ ﻫﺬﻩ ﺍﳊﺎﻟﺔ ﻻ ﺗﺘﺒﻊ ﺗﻮﺯﻳﻊ ‪ t 1a ,T -k‬ﻭ ﺗﺄﺧﺬ‬
‫‪2‬‬
‫ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪dˆ - 1‬‬
‫= ‪td =1‬‬
‫)ˆ‪se(d‬‬
‫ﺣﻴﺚ ˆ‪ d‬ﻫﻮ ﻣﻘﺪﺭ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ‪.‬‬

‫ﻳﺘﻢ ﻗﺒﻮﻝ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ‪ H 0‬ﻭﺗﺄﻛﻴﺪ ﻓﺮﺿﻴﺔ ﻋﺪﻡ ﺍﺳﺘﻘﺮﺍﺭ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﺇﺫﺍ ﻛﺎﻧﺖ ‪ td =1 ñtcv‬ﺃﻱ ﺇﺫﺍ ﻛﺎﻧﺖ ﺍﻹﺣﺼﺎﺀﺓ ﺍﶈﺴﻮﺑﺔ ﺃﻛﱪ ﻣﻦ ﺍﻟﻘﻴﻤﺔ ﺍﳊﺮﺟﺔ‬
‫ﺍﳌﺴﺘﺨﺮﺟﺔ ﻣﻦ ﺟﺪﺍﻭﻝ ﺩﻳﻜﻲ ﻓﻮﻟﺮ‪ .‬ﻭﻛﻤﺎ ﺫﻛﺮﻧﺎ ﺃﻋﻼﻩ ﻓﺈﻥ ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ‬
‫ﲣﺘﻠﻒ ﻣﻦ ﳕﻮﺫﺝ ﺇﱃ ﺁﺧﺮ ﺣﺴﺐ ﺍﳊﺎﻟﺔ ﺍﻷﻭﱃ‪ ،‬ﺍﻟﺜﺎﻧﻴﺔ ﻭﺍﻟﺜﺎﻟﺜﺔ ﻛﻤﺎ ﻳﻈﻬﺮ ﺃﻋﻼﻩ‪.‬‬

‫ﻟﺘﻮﺿﻴﺢ ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ‪ ،‬ﻧﺴﺘﻌﲔ ﺑﺴﻠﺴﻠﺔ ﺯﻣﻨﻴﺔ ﺣﻮﻝ ﺍﻟﻜﺘﻠﺔ ﺍﻟﻨﻘﺪﻳﺔ ‪ M2‬ﻣﻦ‬
‫‪ 1963‬ﺇﱃ ‪ 2005‬ﻭﻧﺮﻳﺪ ﺩﺭﺍﺳﺔ ﺍﺳﺘﻘﺮﺍﺭﻳﺘﻬﺎ ﰒ ﲢﺪﻳﺪ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻷﻧﺴﺐ ﻟﺘﺤﻮﻳﻠﻬﺎ‬
‫ﺇﱃ ﺳﻠﺴﻠﺔ ﻣﺴﺘﻘﺮﺓ‪.‬‬

‫ﺑﺎﻟﺮﺟﻮﻉ ﺇﱃ ﺍﻟﺼﻔﺤﺎﺕ )‪ (113‬ﻭ)‪ (114‬ﻭ )‪ (115‬ﻹﻋﺎﺩﺓ ﳐﺘﻠﻒ ﻋﻤﻠﻴﺎﺕ‬

‫ﺇﺩﺧﺎﻝ ﺍﳌﻌﻄﻴﺎﺕ ﺍﻟﺴﻨﻮﻳﺔ )‪ (Annual‬ﻭﻓﻖ ﺍﻟﻠﻮﺣﺎﺕ ﺍﳌﺪﺭﺟﺔ‪ ،‬ﻣﻊ ﺇﺗﺒﺎﻉ ﺍﳋﻄﻮﺍﺕ‬
‫ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪Quick‬‬ ‫‪Series Statistics‬‬ ‫‪Unit root‬‬ ‫‪Series name‬‬
‫ﺑﻌﺪ ﲢﺪﻳﺪ ﺍﺳﻢ ﺍﻟﺴﻠﺴﻠﺔ‪ ،‬ﻫﻨﺎ ‪ M2‬ﻭﺍﻋﺘﻤﺎﺩﺍﹰ ﻋﻠﻰ ﺍﻟﻨﻤﻮﺫﺝ )‪ (5.57‬ﳝﻜﻦ‬
‫ﺍﻻﻧﻄﻼﻕ ﻣﻦ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﻌﺎﻡ ﻟﻺﺟﺎﺑﺔ ﻋﻠﻰ ﺫﻟﻚ ﺍﻻﻧﺸﻐﺎﻝ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﻟﱪﻧﺎﻣﺞ‬
‫‪..Eviews‬‬

‫‪214‬‬
‫ﺍﻟﻮﺣﺪﺓ‪ :‬ﻣﻠﻴﻮﻥ ﺩ‪.‬ﺝ‪.‬‬ ‫‪.2005‬‬ ‫ﺍﻟﻜﺘﻠﺔ ﺍﻟﻨﻘﺪﻳﺔ ‪ M2‬ﰲ ﺍﳉﺰﺍﺋﺮ ﻣﻦ ‪ 1963‬ﺇﱃ‬ ‫ﺍﳉﺪﻭﻝ )‪:(38‬‬
‫‪t‬‬ ‫اﻟﻣﺷﺎھدة‬ ‫‪t‬‬ ‫اﻟﻣﺷﺎھدة‬ ‫‪t‬‬ ‫اﻟﻣﺷﺎھدة‬
‫‪1‬‬ ‫‪4,724.00‬‬ ‫‪15‬‬ ‫‪51,950.00‬‬ ‫‪29‬‬ ‫‪415,270.00‬‬
‫‪2‬‬ ‫‪4,724.00‬‬ ‫‪16‬‬ ‫‪67,458.00‬‬ ‫‪30‬‬ ‫‪515,902.00‬‬
‫‪3‬‬ ‫‪5,255.00‬‬ ‫‪17‬‬ ‫‪79,688.00‬‬ ‫‪31‬‬ ‫‪627,427.00‬‬
‫‪4‬‬ ‫‪5,778.00‬‬ ‫‪18‬‬ ‫‪93,538.00‬‬ ‫‪32‬‬ ‫‪723,514.00‬‬
‫‪5‬‬ ‫‪7,553.00‬‬ ‫‪19‬‬ ‫‪109,154.00‬‬ ‫‪33‬‬ ‫‪799,562.00‬‬
‫‪6‬‬ ‫‪10,147.00‬‬ ‫‪20‬‬ ‫‪137,889.00‬‬ ‫‪34‬‬ ‫‪915,058.00‬‬
‫‪7‬‬ ‫‪12,120.00‬‬ ‫‪21‬‬ ‫‪165,926.00‬‬ ‫‪35‬‬ ‫‪1,081,518.00‬‬
‫‪8‬‬ ‫‪13,076.00‬‬ ‫‪22‬‬ ‫‪194,718.00‬‬ ‫‪36‬‬ ‫‪1,592,461.00‬‬
‫‪9‬‬ ‫‪13,925.00‬‬ ‫‪23‬‬ ‫‪223,860.00‬‬ ‫‪37‬‬ ‫‪1,789,350.00‬‬
‫‪10‬‬ ‫‪18,139.00‬‬ ‫‪24‬‬ ‫‪227,017.00‬‬ ‫‪38‬‬ ‫‪2,022,534.00‬‬
‫‪11‬‬ ‫‪22,930.00‬‬ ‫‪25‬‬ ‫‪257,896.00‬‬ ‫‪39‬‬ ‫‪2,473,516.00‬‬
‫‪12‬‬ ‫‪25,772.00‬‬ ‫‪26‬‬ ‫‪292,963.00‬‬ ‫‪40‬‬ ‫‪2,901,532.00‬‬
‫‪13‬‬ ‫‪33,749.00‬‬ ‫‪27‬‬ ‫‪308,146.00‬‬ ‫‪41‬‬ ‫‪3,354,422.00‬‬
‫‪14‬‬ ‫‪43,605.00‬‬ ‫‪28‬‬ ‫‪343,005.00‬‬ ‫‪42‬‬ ‫‪3,738,037.00‬‬
‫‪43‬‬ ‫‪4,146,906.00‬‬
‫ﺍﳌﺼﺪﺭ‪ :‬ﺑﻨﻚ ﺍﳉﺰﺍﺋﺮ‪.‬‬

‫ﺍﻟﺴﻠﺴﻠﺔ ‪.M2‬‬ ‫ﺍﳌﺮﺣﻠﺔ ﺍﻷﻭﱃ‪ :‬ﲢﺪﻳﺪ ﺍﻻﺧﺘﺒﺎﺭ)‪ (unit root‬ﰒ‬

‫ﺍﳌﺮﺣﻠﺔ ﺍﻟﺜﺎﻧﻴﺔ‪ :‬ﺗﻌﻴﲔ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﻌﺎﻡ ﻭﲢﺪﻳﺪ ﺩﺭﺟﺔ ﺗﺄﺧﲑ ﺍﻟﻔﺮﻭﻕ‪.‬‬

‫‪215‬‬
‫) ‪Trend‬‬ ‫ﻟﻘﺪ ﰎ ﺍﺧﺘﻴﺎﺭ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﻌﺎﻡ ﺍﻟﺬﻱ ﳛﺘﻮﻱ ﻛﻞ ﻣﺘﻐﲑﺍﺕ ﺍﻟﻨﻤﻮﺫﺝ‬
‫‪ (and intercept‬ﻟﻼﻧﻄﻼﻕ ﰲ ﻋﻤﻠﻴﺔ ﺗﻄﺒﻴﻖ ﺍﺧﺘﺒﺎﺭ ﺩﻳﻜﻲ ﻓﻮﻟﺮ ﻟﻠﻜﺸﻒ ﻋﻦ‬
‫ﻣﺪﻯ ﺍﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ .‬ﻧﻼﺣﻆ ﺃﻋﻼﻩ ﺃﻧﻨﺎ ﺍﺧﺘﺮﻧﺎ ﺍﺧﺘﺒﺎﺭ ﺩﻳﻜﻲ ﻓﻮﻟﺮ‬
‫ﺍﳌﻄﻮﺭ)‪ (Augmented Dickey-Fuller Test‬ﻣﻊ ﺇﻋﻄﺎﺀ ﺍﻟﺮﻗﻢ )‪ (0‬ﻟﻠﻔﺮﻭﻕ‬
‫ﺍﳌﺆﺧﺮﺓ )‪ (lagged differences‬ﻟﻴﺼﺒﺢ ﺍﺧﺘﺒﺎﺭ ﺩﻳﻜﻲ ﻓﻮﻟﺮ ﺍﻟﺒﺴﻴﻂ‪.‬‬

‫ﺇﻥ ﲢﺪﻳﺪ ﺍﳌﺘﻐﲑﺍﺕ ﰲ ﻣﺴﺘﻮﺍﻫﺎ ﺍﻷﺻﻠﻲ )‪ (levels‬ﻳﺘﻢ ﻗﺒﻞ ﻛﺸﻒ ﻋﺪﻡ‬

‫ﺍﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‪ .‬ﻭﻧﻠﺠﺄ ﻓﻴﻤﺎ ﺑﻌﺪ ﺃﻱ ﰲ ﺣﺎﻟﺔ ﻭﺟﻮﺩ ﺍﳉﺬﺭ ﺍﻷﺣﺎﺩﻱ‬
‫ﺇﱃ ﺍﺧﺘﻴﺎﺭ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ)‪ (1st difference‬ﺃﻭ ﺍﻟﺜﺎﻧﻴﺔ ) ‪2nd‬‬
‫‪ (difference‬ﺇﺫﺍ ﺗﻄﻠﺒﺖ ﺍﻟﺴﻠﺴﻠﺔ ﺫﻟﻚ ﻟﺘﺘﺤﻮﻝ ﺇﱃ ﻣﺴﺘﻘﺮﺓ‪.‬‬

‫‪216‬‬
 ‫ﺍﻟﻤﺼﺩﺭ‪ :‬ﻤﺴﺘﺨﺭﺝ ‪ Eviews‬ﺒﺎﺴﺘﺨﺩﺍﻡ ﺍﻟﺘﻌﻠﻴﻤﺔ ‪. unit root test‬‬

‫ﻳﺘﺒﲔ ﻣﻦ ﻫﺬﻩ ﺍﻟﻨﺘﺎﺋﺞ ﺃﻧﻪ ﻻ ﳝﻜﻦ ﺭﻓﺾ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ﻛﻮﻥ ‪ td =1‬ﺍﶈﺴﻮﺑﺔ‬
‫ﺃﻛﱪ ﻣﻦ ﺍﻟﻘﻴﻤﺔ ﺍﳊﺮﺟﺔ ﻟﻼﺧﺘﺒﺎﺭ ﻋﻨﺪ ﺍﳌﺴﺘﻮﻯ ‪ %5‬ﺍﳌﺘﻔﻖ ﻋﻠﻴﻪ ﺃﻱ‬
‫)‪ td =1 = (3.28)ñ(-3.52‬ﺇﺿﺎﻓﺔ ﺇﱃ ﻏﻴﺎﺏ ﺩﻟﻴﻞ ﻋﻦ ﻭﺟﻮﺩ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﻣﻦ‬
‫ﺧﻼﻝ ﺍﺧﺘﺒﺎﺭ ‪. Durbin-Watson stat.‬ﳑﺎ ﻳﻌﻔﻴﻨﺎ ﻣﻦ ﺍﻟﻠﺠﻮﺀ ﺇﱃ ﺍﺳﺘﺨﺪﺍﻡ ﺍﺧﺘﺒﺎﺭ‬
‫‪. Augmented Dickey-Fuller Test‬‬
‫ﻻ ﳝﻜﻦ ﺭﻏﻢ ﺍﻟﻨﺘﺎﺋﺞ ﺍﻻﳚﺎﺑﻴﺔ ﺍﻟﺴﺎﺑﻘﺔ ﺍﻟﻮﻗﻮﻑ ﻋﻨﺪ ﻫﺬﺍ ﺍﳊﺪ ﻣﻦ ﺍﻟﺒﺤﺚ‬
‫ﻟﺴﺒﺐ ﺑﺴﻴﻂ ﻣﺘﻤﺜﻞ ﰲ ﻋﺪﻡ ﻣﻌﻨﻮﻳﺔ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ‪ .‬ﻟﺬﺍ ﻧﻨﺘﻘﻞ‬
‫ﻣﻦ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺜﺎﻟﺚ ﺇﱃ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺜﺎﱐ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪217‬‬
ADF Test Statistic 12.64145 1% Critical Value* -3.5930
5% Critical Value -2.9320
10% Critical Value -2.6039
*MacKinnon critical values for rejection of hypothesis of a unit root.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(M2)
Method: Least Squares
Date: 11/29/09 Time: 15:51
Sample(adjusted): 1964 2005
Included observations: 42 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
M2(-1) 0.140143 0.011086 12.64145 0.0000
C 12786.42 12591.65 1.015469 0.3160
R-squared 0.799806 Mean dependent var 98623.38
Adjusted R-squared 0.794801 S.D. dependent var 151706.9
S.E. of regression 68721.48 Akaike info criterion 25.15996
Sum squared resid 1.89E+11 Schwarz criterion 25.24271
Log likelihood -526.3591 F-statistic 159.8063
Durbin-Watson stat 1.702924 Prob(F-statistic) 0.000000

‫ﰲ ﻫﺬﻩ ﺍﳌﺮﺓ ﺃﻳﻀﺎﹰ ﻻ ﳝﻜﻦ ﺭﻓﺾ ﻓﺮﺿﻴﺔ ﺍﻟﻌﺪﻡ ﲝﻜﻢ ﺃﻥ‬

‫ ﺇﻻ ﺃﻥ ﺍﲣﺎﺫ ﺃﻱ ﻗﺮﺍﺭ ﻋﻠﻰ ﻫﺬﺍ ﺍﳌﺴﺘﻮﻯ ﻳﻌﺘﱪ ﻏﲑ‬. td =1 = (12.64)ñ (-2.93)
‫ ﳑﺎ ﻳﺴﺘﻮﺟﺐ ﺍﻻﻧﺘﻘﺎﻝ ﺇﱃ‬،‫ﻧﺎﺿﺞ ﺍﻵﻥ ﺑﺴﺒﺐ ﻋﺪﻡ ﻣﻌﻨﻮﻳﺔ ﻣﻌﻠﻤﺔ ﺍﳊﺪ ﺍﻟﺜﺎﺑﺖ‬
:‫ﺍﻟﻨﻤﻮﺫﺝ ﺍﻷﻭﻝ ﻭﻛﻤﺎ ﻳﻠﻲ‬

218
 ADF Test Statistic 15.65536 1% Critical Value* -2.6182
5% Critical Value -1.9488
10% Critical Value -1.6199
*MacKinnon critical values for rejection of hypothesis of a unit root.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(M2)
Sample(adjusted): 1964 2005
Included observations: 42 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
M2(-1) 0.146214 0.009340 15.65536 0.0000
R-squared 0.794645 Mean dependent var 98623.38
Adjusted R-squared 0.794645 S.D. dependent var 151706.9
S.E. of regression 68747.61 Akaike info criterion 25.13779
Sum squared resid 1.94E+11 Schwarz criterion 25.17917
Log likelihood -526.8937 Durbin-Watson stat 1.674363

‫ﲝﻜﻢ ﺃﻥ ﺍﻹﺣﺼﺎﺀﺓ ﺍﶈﺴﻮﺑﺔ ﻗﺪ ﲢﺴﻨﺖ ﻗﻴﻤﺘﻬﺎ ﻭﻫﻲ ﺃﻛﱪ ﻣﻦ ﺍﻟﻘﻴﻤﺔ‬

‫ ﺇﱃ‬M2 ‫ ﻳﺘﻢ ﲢﻮﻳﻞ‬.M2 ‫ ﳝﻜﻦ ﺍﻹﻗﺮﺍﺭ ﺑﻌﺪﻡ ﺍﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ‬،‫ﺪﻭﻟﺔ‬‫ﺍ‬
‫ﺳﻠﺴﻠﺔ ﻣﺴﺘﻘﺮﺓ ﻋﻦ ﻃﺮﻳﻖ ﺍﻋﺘﻤﺎﺩ ﻃﺮﻳﻘﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﻭﺣﺴﺐ‬
.‫ﳕﻮﺝ ﺍﻟﺸﻜﻞ ﺍﻷﻭﻝ )ﳕﻮﺫﺝ ﺍﻻﻧﺘﻘﺎﻝ ﺍﻟﻌﺸﻮﺍﺋﻲ ﺑﺪﻭﻥ ﺣﺪ ﺛﺎﺑﺖ( ﺍﻟﻮﺍﺭﺩ ﺃﻋﻼﻩ‬

‫ﻳﺴﺘﻄﻴﻊ ﺍﻟﻄﺎﻟﺐ ﻭﻋﻨﺪ ﻫﺬﺍ ﺍﳌﺴﺘﻮﻯ ﺍﻟﺮﺟﻮﻉ ﺇﱃ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ‬

.‫ﻟﻠﺘﺄﻛﺪ ﻣﻦ ﻣﺪﻯ ﺗﻄﺎﺑﻖ ﺍﻟﻄﺮﻳﻘﺘﲔ ﰲ ﲢﺪﻳﺪ ﺍﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ‬

219
 ‫ﺍﻟﻔﺼﻞ ﺍﻟﺴﺎﺩﺱ‬

‫ﺍﻟﺘﻨﺒﺆ ﻭﻗﻴﺎﺱ ﺍﻟﺪﻗﺔ‬

‫‪ -1‬ﺍﻟﺘﻨﺒﺆ‪:‬‬
‫‪FORECAST‬‬
‫ﳝﺜﻞ ﺍﻟﺘﻨﺒﺆ ﻋﻤﻠﻴﺔ ﻋﺮﺽ ﺣﺎﱄ ﳌﻌﻠﻮﻣﺎﺕ ﻣﺴﺘﻘﺒﻠﻴﺔ ﺑﺎﺳﺘﺨﺪﺍﻡ ﻣﻌﻄﻴﺎﺕ‬
‫ﻣﺸﺎﻫﺪﺓ ﺗﺎﺭﳜﻴﺔ )‪(observed data‬ﺑﻌﺪ ﺩﺭﺍﺳﺔ ﺳﻠﻮﻛﻬﺎ ﺍﳌﺎﺿﻲ‪ .‬ﻳﺘﻤﺜﻞ ﺍﳍﺪﻑ‬
‫ﻣﻦ ﺍﻟﺘﻨﺒﺆ ﺳﻮﺍﺀ ﻷﻏﺮﺍﺽ ﺑﻴﺪﺍﻏﻮﺟﻴﺔ ﺃﻭ ﻋﻤﻠﻴﺔ ﻫﻮ ﻗﻴﺎﺱ ﻗﻴ‪‬ﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ ﳌﺘﻐﲑ‬
‫ﺩﺍﺧﻠﻲ ‪ ،‬ﻛﻮﻥ ﺍﳍﺪﻑ ﺍﻷﺳﺎﺳﻲ ﺍﳌﻨﺘﻈﺮ ﻣﻦ ﳕﺎﺫﺝ ﺍﻟﺴﻼﺳﻞ ﺍﻟﺰﻣﻨﻴﺔ ﻫﻮ ﲢﻘﻴﻖ‬
‫ﺍﻟﺘﻨﺒﺆﺍﺕ‪ ،‬ﺑﻴﻨﻤﺎ ﻳﻜﻮﻥ ﻫﺪﻑ ﺭﺟﻞ ﺍﻷﻋﻤﺎﻝ ﺍﻟﺘﻨﺒﺌﻲ ﳏﺎﻭﻟﺔ ﻣﻌﺮﻓﺔ ﻣﻘﺪﺍﺭ ﻣﺒﻴﻌﺎﺗﻪ‬
‫ﰲ ﺍﻟﺸﻬﺮﻳﻦ ﺍﻟﻘﺎﺩﻣﲔ ﻣﺜﻼﹰ ﻛﻤﺎ ﻳﻜﻮﻥ ﻫﺪﻑ ﺍﻻﻗﺘﺼﺎﺩﻳﲔ ﺍﻟﻮﺣﺪﻭﻳﲔ ﻣﻦ ﻫﺬﻩ‬
‫ﺍﻟﻌﻤﻠﻴﺔ ﺃﻳﻀﺎﹰ ﺍﻟﺘﻨﺒﺆ ﺑﺎﳌﺘﻐﲑﺍﺕ ﺍﳌﻬﻤﺔ ﺑﺎﻟﻨﺴﺒﺔ ﳍﻢ ﻛﺎﻟﺪﺧﻞ ﺍﻟﻘﻮﻣﻲ‪ ،‬ﺍﻻﺳﺘﻬﻼﻙ‬
‫ﺍﻹﲨﺎﱄ ﻭﻣﺴﺘﻮﻯ ﺍﻟﺘﺸﻐﻴﻞ‪،‬ﺍﱁ‪.‬‬
‫ﺇﻥ ﺍﻟﺘﻨﺒﺆ ﻭﻓﻖ ﻣﻘﺎﺭﺑﺔ ﺍﻟﺒﺎﺏ ﺍﻟﺜﺎﱐ ﻳﺘﻢ ﺑﻌﺪ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﺎﺕ ﳕﻮﺫﺝ‬
‫)‪ ARIMA(p,d,q‬ﺍﻟﺬﻱ ﲡﺎﻭﺯ ﳐﺘﻠﻒ ﻣﺮﺍﺣﻞ ﺍﻻﺧﺘﺒﺎﺭﺍﺕ ﺍﻟﺴﺎﺑﻘﺔ ﻭﺍ‪‬ﺪﺩ‬
‫ﺑﺎﻟﺪﺭﺟﺎﺕ ‪ p‬ﻭ ‪ d‬ﻭ‪ q‬ﺣﻴﺚ ﻗﻴﻤﺔ ﺍﻟﺘﻨﺒﺆ ﺗﺼﺒﺢ ﺛﺎﺑﺘﺔ ﻭ ﻣﺴﺎﻭﻳﺔﹰ ﻟﻮﺳﻂ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﺑﻌﺪ ﺍﻟﻔﺘﺮﺓ ‪ q‬ﰲ ﳕﺎﺫﺝ ﺍﳌﺘﻮﺳﻄﺎﺕ ﺍﳌﺘﺤﺮﻛﺔ‪ .‬ﳝﻜﻦ ﺗﻠﺨﻴﺺ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﰲ‬
‫ﺍﳌﺮﺍﺣﻞ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ -1‬ﻛﺘﺎﺑﺔ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻘﺪ‪‬ﺭ‬
‫) ‪$y t = f ( f$ , q$ , y, e‬‬ ‫)‪(6.1‬‬

‫‪l = 1, 2,... L‬‬ ‫ﺣﻴﺚ‬ ‫‪T+ l‬‬ ‫‪ -2‬ﺗﻌﻮﻳﺾ ) ‪ ( t‬ﺑـ‬

‫‪219‬‬
‫‪ -3‬ﺗﻌﻮﻳﺾ ﻛﻞ ﺍﻟﻘﻴ‪‬ﻢ ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﻟﻠﻤﺘﻐﲑ ﺍﳋﺎﺹ ﺑﺎﻟﻈﺎﻫﺮﺓ ﺍﳌﺪﺭﻭﺳﺔ‬
‫ﺑﺘﻨﺒﺆﺍ‪‬ﺎ ‪ ،‬ﺑﻴﻨﻤﺎ ﻳﺘﻢ ﺗﻌﻮﻳﺾ ﺍﻷﺧﻄﺎﺀ ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﺑﺎﻟﺼﻔﺮ ﻭﺍﳌﺎﺿﻴﺔ )ﺩﺍﺧﻞ ﺍﻟﻌﻴﻨﺔ(‬
‫ﺑﺎﻟﺒﻮﺍﻗﻲ‪ ،‬ﻭﻟﺘﻮﺿﻴﺢ ﺫﻟﻚ ﻧﺪﺭﺝ ﺍﻷﻣﺜﻠﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪ MA(1) -1‬ﺍﻟﺬﻱ ﻳﻜﺘﺐ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪y t = m + qe t - 1 + e t‬‬

‫ﺑﻌﺪ ﺍﻟﺘﻘﺪﻳﺮ ﻭﺗﻌﻮﻳﺾ ﺍﻷﺧﻄﺎﺀ ﺍﳌﺎﺿﻴﺔ ﺑﺎﻟﺒﻮﺍﻗﻲ ﻳﺼﺒﺢ ﺍﻟﻨﻤﻮﺫﺝ‬

‫‪$y t = m$ + q$ e t -1‬‬
‫ﺍﻟﺘﻨﺒﺆ ﻟﻔﺘﺮﺓ ﻫﻮ‬
‫‪$y T+1 = m$ + q$ e T‬‬
‫ﻟﻔﺘﺮﺗﲔ ﺑﻌﺪ ﺗﻌﻮﻳﺾ ﺍﻟﺒﻮﺍﻗﻲ ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﺑﺎﻟﺼﻔﺮ‬
‫‪$y T+ 2 = m$ + q$ e T+1 = m$‬‬

‫ﺃﻳﻦ ﻳﻜﻮﻥ ﺍﻟﺘﻨﺒﺆ ﺛﺎﺑﺘﺎﹰ ﺑﻌﺪ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﻭﻫﻮ ﻣﺴﺎﻭﻳﺎﹰ ﻟـ ‪. m‬‬
‫^‬

‫)‪:MA (2‬‬ ‫‪-2‬‬

‫‪y t = m + q 1 e t -1 + q 2 e t - 2 + e t‬‬
‫ﺑﺎﻟﺘﺸﺎﺑﻪ‬
‫‪$y t = m$ + q$ 1 e t -1 + q$ 2 e t - 2‬‬

‫‪$y T+1 = m$ + q$ 1 e T + q$ 2 e T-1‬‬

‫‪220‬‬
 ‫‪$y T+ 2 = m$ + q$ 1 e T+1 + q$ 2 e T‬‬

‫‪y T+ 2 = m$ + q$ 2 e T‬‬

‫‪$y T+ 3 = m$‬‬

‫ﻛﺬﻟﻚ‪،‬‬ ‫ﻟـ ‪m$‬‬ ‫ﻣﻨﻪ ﻳﻜﻮﻥ ﺍﻟﺘﻨﺒﺆ ﺛﺎﺑﺘﺎﹰ ﺑﻌﺪ ﺍﻟﺪﺭﺟﺔ ﺍﻟﺜﺎﻧﻴﺔ ﻭﻫﻮ ﻣﺴﺎﻭﻳﺎﹰ‬
‫ﻭﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ ‪. $y T+ L = m$ "L > q‬‬

‫‪: AR(1) -1‬ﻳﻜﺘﺐ ﻫﺬﺍ ﺍﻟﻨﻤﻮﺫﺝ ﻛﻤﺎ ﺭﺃﻳﻨﺎ ﺳﺎﺑﻘﺎﹰ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪y t = d + fy t - 1 + e t‬‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻘﺪﺭ ﻫﻮ‬
‫‪$y t = d$ + f$ y t -1‬‬
‫ﺍﻟﺘﻨﺒﺆ ﻟﻔﺘﺮﺓ ﻭﺍﺣﺪﺓ ﻣﻌﻄﻰ ﺑـ‪:‬‬
‫‪$y T+1 = d$ + f$ y T‬‬

‫ﰲ ﺣﺎﻟﺔ ﺍﻟﻔﺘﺮﺗﲔ‪ ،‬ﺣﻴﺚ ﻳﺘﻢ ﺗﻌﻮﻳﺾ ‪ y‬ﺍﳌﺴﺘﻘﺒﻠﻴﺔ ﺑﺎﻟﺘﻨﺒﺆ ﺍﳌﻨﺎﺳﺐ ﻭﻛﻤﺎ‬

‫ﻳﻠﻲ‪:‬‬
‫‪$y T+ 2 = d$ + f$ y T+1‬‬
‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫‪$y T+ L = d$ + f$ y T+ L -1‬‬

‫ﻻ ﻳﻜﻮﻥ ﻟﻠﺘﻨﺒﺆ ﺑﻌﺪ ﺍﻟﻔﺘﺮﺓ ‪ p‬ﻋﻼﻗﺔ ﺳﻮﻯ ﺑﺘﻨﺒﺆ ﺍﻟﻔﺘﺮﺓ ﺍﻟﺴﺎﺑﻘﺔ‪ ،‬ﻟﺬﺍ ﻭﻣﻦ‬
‫ﺍﻷﻓﻀﻞ ﺍﻻﺳﺘﻌﺎﻧﺔ ‪‬ﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ ﻷﻏﺮﺍﺽ ﺍﻟﺘﻨﺒﺆ ﺍﻟﻘﺼﲑ ﺍﻷﺟﻞ ﻓﻘﻂ ‪(short term‬‬
‫)‪.forecasting‬‬

‫‪221‬‬
 ‫ﺍﳌﺮﻛﺒﺔ‪ARIMA(1,1,1) :‬‬ ‫‪ -3‬ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﺨﺘﻠﻄﺔ‬
‫ﺗﻌﺘﱪ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ ﰲ ﻫﺬﺍ ﺍﻟﻨﻮﻉ ﻣﻦ ﺍﻟﻨﻤﺎﺫﺝ ﻏﲑ ﻣﺴﺘﻘﺮﺓ ‪ ،‬ﻭﰎﹼ ﺇﺯﺍﻟﺔ‬
‫ﻫﺬﻩ ﺍﻟﻈﺎﻫﺮﺓ ﻋﻦ ﻃﺮﻳﻖ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ ﳌﺮﺓ ﻭﺍﺣﺪﺓ )‪، (d = 1‬‬
‫ﻭﲟﻨﻬﺠﻴﺔ ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ ﻧﻜﻮﻥ ﻗﺪ ﺃﺑﻌﺪﻧﺎ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻣﻨﻬﺎ‪ ،‬ﻭﻧﺴﻤﻲ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻨﺎﲡﺔ ﻭﺍﻟﱵ ﻗﺪ ﺗﻜﻮﻥ ﺧﺎﻟﻴﺔ ﻣﻦ ﺍﳌﺮﻛﺒﺔ ﺍﳌﺬﻛﻮﺭﺓ ‪ w t‬ﺇﺫﺍﹰ‪:‬‬

‫‪w t = y t - y t -1‬‬

‫ﺑﺎﻋﺘﺒﺎﺭ ﺃﻥ ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺘﺎﱄ ﻧﺘﺎﺝ ﳐﺘﻠﻒ ﺍﳌﺮﺍﺣﻞ ﺍﳌﺬﻛﻮﺭﺓ ﺳﺎﺑﻘﺎﹰ ﰲ ﳎﺎﻝ‬

‫ﲢﺪﻳﺪ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺘﻨﺒﺆﻳﺔ ﻭ ﺍﳉﺎﻫﺰﺓ ﻟﻼﺳﺘﺨﺪﺍﻡ‬
‫‪$ t = d$ + f$ w t -1 + q$ e t -1‬‬
‫‪w‬‬

‫ﺃﻋﻼﻩ ﳓﺼﻞ ﻋﻠﻰ‬ ‫‪T+ 1‬‬ ‫ﺑﺘﻌﻮﻳﺾ) ‪ ( t‬ﺑـ‬

‫‪$ T+1 = d$ + f$ w T + q$ e T‬‬
‫‪w‬‬
‫ﺣﻴﺚ ‪ e T‬ﲤﺜﻞ ﺁﺧﺮ ﻣﺸﺎﻫﺪﺓ ﻣﻦ ﺷﻌﺎﻉ ﺍﻟﺒﻮﺍﻗﻲ ﺫﻭ ﺍﻟﺒﻌﺪ )‪. ( T ´ 1‬‬
‫ﺇﻥ ﺍﻟﺘﻨﺒﺆ ﻟﻔﺘﺮﺓ ﺇﺿﺎﻓﻴﺔ ﻳﻌﻄﻲ ﺑـ‪:‬‬
‫‪$ T+ 2 = d$ + f$ w T+1 + q$ e T+1‬‬
‫‪w‬‬

‫‪= d$ + fw‬‬
‫‪$‬‬
‫‪T +1‬‬

‫‪e T+ 1 = 0‬‬ ‫ﺣﻴﺚ‬

‫ﺇﻻ ﺃﻧﻪ ﰲ ﺍﻟﻌﻤﻠﻴﺔ ﻫﺬﻩ ﻻ ﳓﺘﺎﺝ ﻟﺘﻨﺒﺆ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳋﺎﻟﻴﺔ ﻣﻦ ﺍﳌﺮﻛﺒﺔ ﺍﳌﻨﺰﻭﻋﺔ‬
‫ﺑﻘﺪﺭ ﻣﺎ ﳓﺘﺎﺝ ﺇﱃ ﺍﻟﺘﻨﺒﺆ ﺍﻟﻜﻠﻲ ﺍﻟﻨﻬﺎﺋﻲ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﳌﺪﺭﻭﺳﺔ‪ ،‬ﻟﺬﺍ ﻳﺘﻢ‬
‫ﺗﻌﻮﻳﺾ ) ‪ ( t‬ﺑـ ‪ T + 1‬ﰒ ‪ T + 2‬ﰲ ﻣﻌﺎﺩﻟﺔ ﺍﻟﻔﺮﻭﻕ ﻣﻦ ﺍﻟﺪﺭﺟﺔ ﺍﻷﻭﱃ‬

‫‪222‬‬
‫ﺍﳌﺬﻛﻮﺭﺓ ﺃﻋﻼﻩ ﻭ‪‬ﺬﺍ ﻧﻜﻮﻥ ﻗﺪ ﺭﺟﻌﻨﺎ ﻣﺮﻛﺒﺔ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﺇﱃ ﺍﻟﺴﻠﺴﻠﺔ‬
‫ﻛﻤﺎ ﻳﻠﻲ‪:‬‬
‫‪w‬‬
‫‪$‬‬ ‫‪T+ 1‬‬ ‫‪= $y T + 1 - y T‬‬
‫ﻭﻣﻨﻪ‬
‫‪$y T + 1 = w‬‬
‫‪$‬‬ ‫‪T+ 1‬‬ ‫‪+yT‬‬

‫‪$y T + 2 = w‬‬
‫‪$ T + 2 + $y T + 1‬‬
‫ﻋﻠﻰ ﺍﻟﻌﻤﻮﻡ‬
‫‪$y T + L = w‬‬
‫‪$ T+ L + y T + L - 1‬‬

‫ﻣﻦ ﺍﻟﻌﻼﻗﺔ ﺍﻷﺧﲑﺓ ﻧﻜﺮﺭ ﻣﻼﺣﻈﺔ ﺃﻥ ﺍﻟﺮﻏﺒﺔ ﰲ ﺍﺳﺘﻌﻤﺎﻝ ﻫﺬﻩ ﺍﻟﻨﻤﺎﺫﺝ‬

‫ﺗﺰﻳﺪ ﰲ ﺣﺎﻟﺔ ﺍﻟﺘﻨﺒﺆ ﺍﻟﻘﺼﲑ ﺍﻷﺟﻞ‪.‬‬
‫‪:GB13‬‬ ‫ﺗﻄﺒﻴﻖ ﻋﻠﻰ‬
‫ﻳﺴﺘﻌﻤﻞ ﰲ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻨﺒﺆ ﺍﻟﻨﻤﻮﺫﺝ )‪ (1‬ﺍﳌﻔﻀﻞ‬
‫‪$y t = 373151 . 28 + 0. 66 y t - 1 + 0. 87e t - 12‬‬

‫ﻭ ‪... T+2‬ﺇﱁ‬ ‫‪T+1‬‬ ‫ﺑﺘﻌﻮﻳﺾ ‪ t‬ﺑـ‬

‫‪$y T + 1 = 373151 . 28 + 0. 66 y T + 0. 87e T - 11‬‬

‫ﺗﻜﻮﻥ ‪ yT‬ﻫﻲ ﺁﺧﺮ ﻣﺸﺎﻫﺪﺓ ﰲ ﺍﻟﺴﻠﺴﻠﺔ ‪ GB13‬ﺇﺫﺍﹰ‬

‫‪$y T+ 1 = 373151 . 28 + 0. 66 y 62 + 0. 87e 51‬‬

‫) ‪yˆ T +1 = 373151 . 28 + 0 . 66 ( 514094 ) + 0 . 87 ( 9762 . 187‬‬

‫‪= 720926 .42‬‬

‫‪223‬‬
‫‪.‬‬ ‫ﺣﻴﺚ ﰎﹰ ﺗﻌﻮﻳﺾ ‪ e 51‬ﻣﻦ ﺷﻌﺎﻉ ﺍﻟﺒﻮﺍﻗﻲ ﺍﳌﻮﻓﺮ ﻣﻦ ﻃﺮﻑ ﺑﺮﻧﺎﻣﺞ ﺍﻟﺘﻘﺪﻳﺮ‬

‫‪ -2‬ﻗﻴﺎﺱ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ‬

‫ﺗﻌﺘﱪ ﻫﺬﻩ ﺍﳌﺮﺣﻠﺔ ﺫﺍﺕ ﺃﳘﻴﺔ ﰲ ﺗﻘﻴ‪‬ﻴﻢ ﺍﻟﻨﻤﻮﺫﺝ ﻟﻸﻏﺮﺍﺽ ﺍﻟﺘﻨﺒﺆﻳﺔ ‪ ،‬ﻭﻣﻦ‬
‫ﺍﳌﻌﺎﻳﲑ ﺍﳌﺴﺘﺨﺪﻣﺔ ﰲ ﻫﺬﺍ ﺍ‪‬ﺎﻝ‪:‬‬

‫‪ -1‬ﻣﺘﻮﺳﻂ ﺍﳋﻄﺄ‪ :‬ﻭﻫﻮ ﻳﻌﱪ ﻋﻠﻰ ﻣﺘﻮﺳﻂ ﺍﻟﻔﺮﻕ ﺑﲔ ﺍﳌﺸﺎﻫﺪﺓ ﻭﺍﻟﺘﻨﺒﺆ‬

‫ﻟﻨﻔﺲ ﺍﻟﻔﺘﺮﺓ ﺍﻟﺰﻣﻨﻴﺔ‪ ،‬ﻭﻳﻌﻄﻰ ﺭﻳﺎﺿﻴ‪‬ﺎ ﰲ ﺍﻟﺸﻜﻞ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪1 T‬‬
‫= ‪ME‬‬ ‫‪å yt - yt‬‬
‫‪p‬‬

‫‪T t =1‬‬
‫ﺣﻴﺚ ‪ yt‬ﺗﻌﱪ ﻋﻠﻰ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﺪﺭﻭﺳﺔ ﺑﻴﻨﻤﺎ ‪ y tp‬ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﻤﻬﺪﺓ ﺃﻭ‬
‫ﺍﳌﹲﺘﻨﺒﺄ ‪‬ﺎ ﺩﺍﺧﻞ ﺍﻟﻌﻴﻨﺔ ﺍﳌﺪﺭﻭﺳﺔ‪.‬‬
‫ﳝﻜﻦ ﺃﻥ ﻳ‪‬ﺄﺧﺬ ﻫﺬﺍ ﺍﳌﻌﻴﺎﺭ ﰲ ﺷﻜﻞﹴ ﻧﺴﱯ ﻭﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪1 T é yt - ytp ù‬‬

‫= ‪PME‬‬ ‫‪åê‬‬
‫‪T t =1 ë yt û‬‬
‫‪ú‬‬

‫ﺇﻻ ﺃﻧﻪ ﻳﻌﺎﺏ ﻋﻠﻰ ﻫﺬﻳﻦ ﺍﳌﻌﻴﺎﺭﻳﻦ ﺃ‪‬ﻤﺎ ﻗﺪ ﻳﻌﻄﻴﺎﻥ ﻧﻈﺮﺓ ﻣﻐﻠﻄﺔ ﻋﻠﻰ ﻣﺪﻯ‬
‫ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ ﻛﻮ‪‬ﻤﺎ ﻳﺘﺠﺎﻫﻼﻥ ﻓﻜﺮﺓ ﺃﻥ ﺍﻷﺧﻄﺎﺀ ﺍﻟﻜﺒﲑﺓ ﺍﳌﻮﺟﺒﺔ ﺗﻠﻐﻲ ﺍﻟﺴﺎﻟﺒﺔ‪،‬‬
‫ﻛﻤﺎ ﺃﻥ ﳏﺎﻭﻟﺔ ﺗﻔﺎﺩﻱ ﻣﺸﻜﻠﺔ ﺍﻹﺷﺎﺭﺓ ﻋﻦ ﻃﺮﻳﻖ ﺍﻟﻘﻴ‪‬ﻢ ﺍﳌﻄﻠﻘﺔ ﻻ ﻳﻔﻲ ﺑﺎﻟﻐﺮﺽ‬
‫ﻛﻮﻥ ﺍﳋﻂ ﺍﻟﺬﻱ ﻳﻀﻤﻦ ﺃﻗﻞ ﻗﻴﻤﺔ ﳍﺬﺍ ﺍﳌﻌﻴﺎﺭ ﻻ ﳝﺜﻞ ﺑﺎﻟﻀﺮﻭﺭﺓ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ‬
‫ﻟﻠﺒﻴﺎﻧﺎﺕ ﺑﺸﻜﻞ ﺟﻴﺪ‪. 1‬‬

‫‪33‬‬ ‫‪ -1‬ﻣﻘﺪﻣﺔ ﰲ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ‪ ،‬ﺹ‬

‫‪224‬‬
‫‪ -2‬ﺟﺬﺭ ﻣﺘﻮﺳﻂ ﻣﺮﺑﻌﺎﺕ ﺍﻟﺒﻮﺍﻗﻲ‪ :‬ﺟﺎﺀ ﻫﺬﺍ ﺍﳌﻌﻴﺎﺭ ﻛﺒﺪﻳﻞ ﻟﻠﺴﺎﺑﻖ ﻭ‬
‫ﺍﻟﺬﻱ ﻳﺮﻣﺰ ﻟﻪ ﻻﺗﻴﻨﻴﺎﹰ ﺑـ ‪ 2 RMSE‬ﻭ ﻫﻮ ﻣﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬
‫‪1 T‬‬
‫= ‪RMSE‬‬ ‫‪å‬‬
‫‪T t =1‬‬
‫‪( yt - ytp ) 2‬‬

‫ﳝﻜﻦ ﺍﻟﺘﻌﺒﲑ ﻋﻨﻪ ﺃﻳﻀﺎﹰ ﰲ ﺷﻜﻞ ﻧﺴﱯ ﻛﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪1 T é ( yt - ytp ) 2 ù‬‬
‫= ‪PRMSE‬‬ ‫‪åê y ú‬‬
‫‪T t =1 ë‬‬ ‫‪t‬‬ ‫‪û‬‬

‫ﻳﺘﻤﻴﺰ ﻫﺬﻳﻦ ﺍﻷﺧﲑﻳﻦ ﻋﻦ ﺳﺎﺑﻘﻴﻬﻤﺎ ﰲ ﺃ‪‬ﻤﺎ ﻻ ﻳﻔﺮﻗﺎﻥ ﺑﲔ ﺍﻟﺒﻮﺍﻗﻲ‬

‫)ﺍﻟﻔﻮﺍﺭﻕ( ﺳﻮﺍﺀ ﻛﺎﻧﺖ ﻣﻮﺟﺒﺔ ﺃﻭ ﺳﺎﻟﺒﺔ‪ ،‬ﻭﺑﺎﻟﺘﺎﱄ ﰎﹼ ﺇﺑﻌﺎﺩ ﻓﻜﺮﺓ ﺍﻷﺧﻄﺎﺀ ﺍﳌﻮﺟﺒﺔ‬
‫ﺍﻟﱵ ﺗﻠﻐﻲ ﺍﻟﺴﺎﻟﺒﺔ‪.‬‬

‫ﻳﻌﺘﱪ ‪ RMSE‬ﻣﻦ ﺃﻫﻢ ﺍﳌﻌﺎﻳﲑ ﺍﳌﺴﺘﺨﺪﻣﺔ ﰲ ﺍﳌﻔﺎﺿﻠﺔ ﺑﲔ ﳎﻤﻮﻋﺔ ﻣﻦ‬

‫ﺍﻟﻨﻤﺎﺫﺝ ﻋﻠﻰ ﺃﺳﺎﺱ ﺃﺻﻐﺮ ‪.RMSE‬‬

‫‪ -3‬ﻣﻘﻴﺎﺱ ﺍﻻﳓﺪﺍﺭ ﻭﺍﻻﺭﺗﺒﺎﻁ‪ :‬ﻭ ﻫﻮ ﻳﻌﺘﻤﺪ ﻋﻠﻰ ﺇﺟﺮﺍﺀ ﺍﳓﺪﺍﺭ ﺑﲔ‬

‫ﺍﳌﺸﺎﻫﺪﺍﺕ ﺍﳊﺎﻟﻴﺔ ﻭ ﺍﻟﺘﻨﺒﺆ ﲟﻌﲎ‪:‬‬

‫‪y t = b 0 + b 1 y pt + e t‬‬

‫ﻳﺘﻢ ﺗﻘﺪﻳﺮ ﻣﻌﻠﻤﱵ ﺍﻟﻨﻤﻮﺫﺝ ﺑﻄﺮﻳﻘﺔ ﺍﳌﺮﺑﻌﺎﺕ ﺍﻟﺼﻐﺮﻯ‪ ،‬ﻭﻳﻜﻮﻥ ﺍﻟﺘﻨﺒﺆ ﻓﻌﺎ ﹰﻻ‬
‫ﺇﺫﺍ ﻛﺎﻧﺖ ‪ b$ 1 , b$ 0‬ﻻ ﲣﺘﻠﻔﺎﻥ ﻣﻌﻨﻮﻳﺎﹰ ﻋﻦ ﺍﻟﺼﻔﺮ )‪ 0‬ﻭﺍﻟﻮﺍﺣﺪ )‪ (1‬ﻋﻠﻰ ﺍﻟﺘﺮﺗﻴﺐ‬
‫ﺑﺎﺳﺘﺨﺪﺍﻡ ﺇﺣﺼﺎﺀﺓ ﻣﻨﺎﺳﺒﺔ‪.‬‬

‫(‬
‫‪2- Root Mean Square Error‬‬

‫‪225‬‬
‫ﻳﻨﺘﻘﺪ ﻫﺬﺍ ﺍﳌﻌﻴﺎﺭ ﺃﻳﻀﺎﹰ ﻋﻠﻰ ﺃﺳﺎﺱ ﺃﻧﻪ ﻻ ﻳﻨﻈﺮ ﺇﻻ ﺇﱃ ﺟﻮﺩﺓ ﺍﻟﻌﻼﻗﺔ ﺑﲔ‬
‫ﺍﳌﺸﺎﻫﺪﺓ ﻭﺍﻟﺘﻨﺒﺆ ﻭﻳﻬﻤﻞ ﺣﺠﻢ ﻭﺳﻠﻮﻙ ﺍﳋﻄﺄ ﺍﻟﺘﻨﺒﺌﻲ ﻭﺑﺎﻟﺘﺎﱄ ﻓﺈﻧﻪ ﺳﻴﻌﻄﻲ ﻧﻈﺮﺓ‬
‫ﻣﻐﻠﻄﺔ ﺣﻮﻝ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ‪.‬‬

‫‪ -4‬ﻣﻌﻴﺎﺭ ﺛﺎﻳﻞ ‪ :Theil's U statistic‬ﻭﻫﻮ ﻣﻌﻄﻰ ﰲ ﺍﻟﺼﻴﻐﺔ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪RMSE‬‬
‫=‪U‬‬
‫‪1 T‬‬ ‫‪1 T‬‬
‫‪å( y t )2 +‬‬ ‫‪å ( y pt ) 2‬‬
‫‪T t =1‬‬ ‫‪T t =1‬‬

‫‪،‬‬ ‫‪U= 1‬‬ ‫ﻓﻴﻜﻮﻥ ﺍﻟﺘﻨﺒﺆ ﺟﻴﺪﺍﹰ ﳌﺎ ﻳﻜﻮﻥ ‪ ، U = 0‬ﻭﺗﻜﻮﻥ ﺍﻟﻌﻤﻠﻴﺔ ﻓﺎﺷﻠﺔ ﳌﺎ‬
‫ﻭﻋﻤﻠﻴﺎﹰ ﻳﺘﺬﺑﺬﺏ ﻫﺬﺍ ﺍﳌﻘﻴﺎﺱ ﺑﲔ ﻫﺎﺗﲔ ﺍﻟﻘﻴﻤﺘﲔ ‪.‬‬

‫‪ -5‬ﺍﻗﺘﻔﺎﺀ ﺍﻷﺛﺮ )‪ :(Tracking performance‬ﺗﺘﻤﺜﻞ ﻫﺬﻩ ﺍﻟﻄﺮﻳﻘﺔ ﰲ‬

‫ﻗﻴﺎﺱ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺆ ﻣﻦ ﺧﻼﻝ ﻣﺪﻯ ﻗﺪﺭﺗﻪ ﰲ ﺍﻗﺘﻔﺎﺀ ﺃﺛﺮ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻷﺻﻠﻴﺔ ﻭﺗﺘﺒﻊ ﻧﻘﺎﻁ‬
‫ﺍﻧﻌﻄﺎﻓﻬﺎ ﺑﺮﺷﺎﻗﺔ ﻛﻤﺎ ﺫﻛﺮﻧﺎ ﰲ ﺍﻟﺒﺎﺏ ﺍﻷﻭﻝ‪ ،‬ﻭﻟﺘﻮﺿﻴﺢ ﻫﺬﻩ ﺍﻟﻌﻤﻠﻴﺔ ﻧﺴﺘﻌﲔ‬
‫ﺩﺍﺋﻤﺎﹰ ﺑﺎﻟﺮﺳﻮﻣﺎﺕ ﺍﻟﺒﻴﺎﻧﻴﺔ ﻟﻠﺴﻠﺴﻠﺘﲔ ﺍﻷﺻﻠﻴﺔ ﻭﺍﻟﺘﻨﺒﺆﻳﺔ‪.‬‬
‫ﻧﺮﻛﺰ ﰲ ﻫﺬﺍ ﺍ‪‬ﺎﻝ ﻋﻠﻰ ﻣﺪﻯ ﺍﺳﺘﻄﺎﻋﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﺍﻟﻨﺎﲡﺔ ﺳﻮﺍﺀً ﻣﻦ‬
‫ﻋﻤﻠﻴﺎﺕ ﺍﻟﺘﻤﻬﻴﺪ‪ ،‬ﺍﻻﳓﺪﺍﺭ ﻭﺃﺛﻨﺎﺀ ﺍﻟﻘﻴﺎﻡ ﺑﻌﻤﻠﻴﺎﺕ ﺍﻟﺘﻨﺒﺆ ﺍﻟﺘﺎﺭﳜﻲ‪ ،‬ﻣﻦ ﺗﺘﺒﻊ ﺧﺎﺻﺔ‬
‫ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻟﻠﻤﻨﺤﲎ ﻭﻛﺬﺍ ﺗﺴﺠﻴﻞ ﻭ ﺗﺘﺒﻊ ﳐﺘﻠﻒ ﻧﻘﺎﻁ ﺍﻻﻧﻌﻄﺎﻑ‪.‬‬
‫ﻧﺪﺭﺝ ﻛﻤﺜﺎﻝ ﻋﻠﻰ ﺫﻟﻚ ﺍﻟﺪﺭﺍﺳﺔ ﺍﻟﺘﺎﻟﻴﺔ ﺍﳌﺘﻌﻠﻘﺔ ﺑﺴﻠﺴﻠﺔ ﺯﻣﻨﻴﺔ ﺣﻮﻝ ﺇﻧﺘﺎﺝ‬
‫ﺍﻟﺴﻤﻚ‪ ،‬ﺣﻴﺚ ﺍﺳﺘﻄﺎﻋﺖ ﺍﻟﺴﻠﺴﻠﺔ ﺍﳌﻤﻬﺪﺓ ‪ ys‬ﻣﻦ ﺍﻗﺘﻔﺎﺀ ﺃﺛﺮ ﺍﻟﺴﻠﺴﻠﺔ ‪ y‬ﺑﺸﻜﻞ‬
‫ﻣﻠﻔﺖ ﻟﻼﻧﺘﺒﺎﻩ ﻣﻦ ﺣﻴﺚ ﺗﺘﺒﻊ ﺳﻠﻮﻙ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﻛﺬﺍ ﻧﻘﺎﻁ ﺍﻻﻧﻌﻄﺎﻑ ﺍﻟﱵ‬
‫ﺗﻌﺘﱪ ﻣﻬﻤﺔ ﻣﻦ ﺣﻴﺚ ﻗﻴﺎﺱ ﺃﺩﺍﺀ ﺍﻟﻨﻤﻮﺫﺝ‪.‬‬

‫‪226‬‬
 ‫ﺍﻟﺸﻜﻞ )‪ :(38‬ﺍﻗﺘﻔﺎﺀ ﺍﻷﺛﺮ‬
‫‪2000‬‬

‫‪1500‬‬

‫‪1000‬‬

‫‪500‬‬

‫‪0‬‬
‫‪71‬‬ ‫‪72‬‬ ‫‪73‬‬ ‫‪74‬‬ ‫‪75‬‬ ‫‪76‬‬ ‫‪77‬‬ ‫‪78‬‬ ‫‪79‬‬

‫‪Y‬‬ ‫‪YS‬‬

‫‪227‬‬
 ‫ﲤﺮﻳﻦ ‪:1‬‬
‫‪ -‬ﺃﺣﺴﺐ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻨﻈﺮﻳﺔ ﻟﻠﻨﻤﻮﺫﺝ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪y t = f 4 y t -4 + e t‬‬
‫‪t = 1, 32‬‬
‫ﰒ ﻣﺜﻠﻬﺎ ﺑﻴﺎﻧﻴﺎﹰ ‪.‬‬

‫ﲤﺮﻳﻦ ‪:2‬‬
‫‪ -‬ﺃﺣﺴﺐ ﻣﻌﺎﻣﻼﺕ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻨﻈﺮﻳﺔ ﻟﻠﻨﻤﻮﺫﺝ ﺍﻟﺘﺎﱄ‪:‬‬

‫‪y t = q 4 e t -4 + e t‬‬
‫‪t = 1, 32‬‬
‫ﻓﻤﺜﻠﻬﺎ ﺑﻴﺎﻧﻴﺎﹰ ﺃﻳﻀﺎﹰ‪.‬‬

‫ﲤﺮﻳﻦ ‪:3‬‬
‫‪ -‬ﻣﺎﺫﺍ ﺗﺴﺘﻨﺘﺞ ﻣﻦ ﺍﻟﺪﺍﻟﺘﲔ ﺍﻟﻮﺍﺭﺩﺗﲔ ﰲ ﺍﻟﺴﺆﺍﻟﲔ ﺍﻷﻭﻝ ﻭﺍﻟﺜﺎﱐ ﺃﻋﻼﻩ ‪.‬‬

‫ﲤﺮﻳﻦ‬ ‫‪:4‬‬
‫ﺍ‪ -‬ﺃﺛﺒﺖ ﺃﻥ ﺗﻨﺒﺆ ﳕﻮﺫﺝ )‪ ARIMA(0,1,1‬ﺍﻟﺘﺎﱄ ﻳﻌﻄﻲ ﻧﻔﺲ ﺗﻨﺒﺆ ﳕﻮﺫﺝ‬
‫ﺍﻟﺘﻤﻬﻴﺪ ﺍﻷﺳﻲ ﺍﻷﺣﺎﺩﻱ )‪.(SES‬‬
‫‪y t - y t - 1 = e t + qe t - 1‬‬

‫ﺏ‪ -‬ﺍﻋﺘﻤﺎﺩﺍ ﻋﻠﻰ )ﺍ( ﺃﺣﺴﺐ ﺑﻄﺮﻳﻘﺔ )‪ (SES‬ﺍﻟﺘﻨﺒﺆ ‪. $y T+ 2‬‬

‫‪228‬‬
 ‫ﲤﺮﻳﻦ ‪:5‬‬
‫‪ -‬ﺑﲔ ﺑﺎﻟﺘﻔﺼﻴﻞ ﻛﻴﻔﻴﺔ ﺗﻘﺪﻳﺮ ﳕﻮﺫﺝ )‪ MA(1‬ﺑﻄﺮﻳﻘﺔ‪:‬‬
‫ﺍ‪ -‬ﺍﻟﺒﺤﺚ ﺍﻟﺸﺒﻜﻲ )‪.(Grid search‬‬
‫ﺑـ‪ -‬ﻏﻮﺱ ‪ -‬ﻧﻴﻮﺗﻦ ﺍﻟﺘﻜﺮﺍﺭﻳﺔ‪.‬‬
‫ﲤﺮﻳﻦ ‪:6‬‬
‫‪ -‬ﺑﺎﺳﺘﻌﻤﺎﻝ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﻟﻮﺍﺭﺩﺓ ﺃﺩﻧﺎﻩ ﺍﳌﻤﺜﻠﺔ ﻟﺪﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ )‪ (AC‬ﻭﺩﺍﻟﺔ‬
‫ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﳉﺰﺋﻴﺔ )‪ (PAC‬ﺍﳋﺎﺻﺔ ﲟﺘﻐﲑ ﺳﻠﻌﺔ ﻣﻌﻴﻨﺔ ﺍﳌﻄﻠﻮﺏ‪:‬‬
‫ﺍ‪ -‬ﺍﺳﺘﻨﺘﺎﺝ ﺍﳌﻌﻠﻮﻣﺎﺕ ﺍﻷﺳﺎﺳﻴﺔ ﻣﻨﻬﻤﺎ ﻣﻊ ﻛﺘﺎﺑﺔ ﻛﻞ ﺍﻷﺩﻭﺍﺕ ﺍﻹﺣﺼﺎﺋﻴﺔ‬
‫ﺍﳌﺴﺘﻌﻤﻠﺔ ‪.‬‬
‫ﺏ‪ -‬ﺣﺪﺩ ﺍﻟﻨﻤﻮﺫﺝ ﺍﳌﻨﺎﺳﺐ ‪.‬‬
‫ﺟـ‪ -‬ﻗﺪﺭ ﺭﻗﻤﻴﺎﹰ ﳌﻌﻠﻤﺔ ﺃﻭ ﻣﻌﻠﻤﺎﺕ ﺍﻟﻨﻤﻮﺫﺝ ﺑﻄﺮﻳﻘﺔ ﻣﻨﺎﺳﺒﺔ ﻭﺑﺴﻴﻄﺔ ﺇﻥ‬
‫ﺃﻣﻜﻦ‪.‬‬

‫‪PAC‬‬ ‫‪AC‬‬ ‫‪k‬‬ ‫‪PAC‬‬ ‫‪AC‬‬ ‫‪k‬‬

‫‪-0.163‬‬ ‫‪0.192‬‬ ‫‪8‬‬ ‫‪0.731‬‬ ‫‪0.731‬‬ ‫‪1‬‬
‫‪0.031‬‬ ‫‪0.14‬‬ ‫‪9‬‬ ‫‪0.014‬‬ ‫‪0.541‬‬ ‫‪2‬‬
‫‪0.00‬‬ ‫‪0.11‬‬ ‫‪10‬‬ ‫‪0.064‬‬ ‫‪0.43‬‬ ‫‪3‬‬
‫‪0.00‬‬ ‫‪0.08‬‬ ‫‪11‬‬ ‫‪0.031‬‬ ‫‪0.354‬‬ ‫‪4‬‬
‫‪-0.137‬‬ ‫‪0.01‬‬ ‫‪12‬‬ ‫‪0.025‬‬ ‫‪0.297‬‬ ‫‪5‬‬
‫‪0.073‬‬ ‫‪0.02‬‬ ‫‪13‬‬ ‫‪0.14‬‬ ‫‪0.311‬‬ ‫‪6‬‬
‫‪-0.033‬‬ ‫‪0.002‬‬ ‫‪14‬‬ ‫‪0.009‬‬ ‫‪0.298‬‬ ‫‪7‬‬

‫ﲤﺮﻳﻦ ‪:7‬‬
‫‪ -‬ﺃﺭﺍﺩ ﺑﺎﺣﺚ ﰲ ﻣﻴﺪﺍﻥ ﺍﻷﺳﻌﺎﺭ ﺩﺭﺍﺳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ‪ p t‬ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ‬
‫‪ 95.12-85.1‬ﲟﺠﻤﻮﻉ )‪ (132‬ﻣﺸﺎﻫﺪﺓ‪ ،‬ﻭﺑﻌﺪ ﺗﻔﻘﺪﻩ ﳌﺮﻛﺒﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ‬
‫ﺍﳌﺘﻤﺜﻠﺔ ﰲ ﺍﻻﲡﺎﻩ ﺍﻟﻌﺎﻡ ﻭﺍﻟﻔﺼﻠﻴﺔ ﻭﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﻗﺮﺭ ﺍﻟﻘﻴﺎﻡ ﲟﺎ ﻳﻠﻲ‪:‬‬

‫‪229‬‬
‫ﻟﻠﺴﻠﺴﻠﺔ ‪p t‬‬ ‫ﺍ‪ -‬ﺣﺴﺎﺏ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻜﻠﻴﺔ ‪ ac‬ﻭﺍﳉﺰﺋﻴﺔ‬
‫‪pac‬‬
‫)ﺍﻟﺸﻜﻞ‪.(f1‬‬
‫ﺑـ‪ -‬ﺣﺴﺎﺏ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻜﻠﻴﺔ ‪ ac‬ﻭﺍﳉﺰﺋﻴﺔ ‪pac‬‬
‫ﻟﻠﺴﻠﺴﻠﺔ ‪ (f2) Dp t‬ﺣﻴﺚ ‪:‬‬
‫‪Dp t = p t - p t -1‬‬

‫ﺟـ‪ -‬ﺣﺴﺎﺏ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻜﻠﻴﺔ ‪ ac‬ﻭﺍﳉﺰﺋﻴﺔ ‪ pac‬ﻟﻠﺴﻠﺴﻠﺔ‬

‫‪ ( f3) D12 p t‬ﺣﻴﺚ‪:‬‬
‫‪D12 p t = p t - p t -12‬‬
‫ﺩ ‪ -‬ﺣﺴﺎﺏ ﺩﺍﻟﺔ ﺍﻻﺭﺗﺒﺎﻁ ﺍﻟﺬﺍﰐ ﺍﻟﻜﻠﻴﺔ ‪ ac‬ﻭﺍﳉﺰﺋﻴﺔ ‪ pac‬ﻟﻠﺴﻠﺴﻠﺔ‬
‫‪ DD12 p t‬ﺍﻟﻮﺍﺭﺩﺓ ﰲ ﺍﻟﺸﻜﻞ )‪(f4‬ﺣﻴﺚ‪:‬‬
‫‪DD12 p t = D12 p t - D12 p t - 1‬‬

‫ﻭﺗﻮﺻﻞ ﰲ ﺍﻷﺧﲑ ﺇﱃ ﺍﻟﻨﻤﺎﺫﺝ ﺍﳌﻘﺪﺭﺓ ﺍﻟﺘﺎﻟﻴﺔ ‪:‬‬

‫‪:( DD 12 p t‬‬ ‫‪= wt‬‬ ‫ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﺴﻠﺴﻠﺔ ‪) DD12 p t‬‬
‫}‪{1‬‬
‫‪w t =. 007778 + e t + -0. 628104 e t - 1‬‬
‫) ‪( 0. 08277‬‬ ‫) ‪( 0. 07188‬‬
‫‪RSS = 670 S = 2. 39 T = 119‬‬

‫}‪{2‬‬
‫‪w t =. 0936 + 0. 966 w t - 1 + e t + -0. 599 e t -1‬‬
‫) ‪( 0. 028‬‬ ‫) ‪( 0. 085‬‬
‫‪RSS = 660 S 2 = 5. 64 T = 119‬‬

‫‪:( D12 p t‬‬ ‫‪=z‬‬ ‫‪t‬‬ ‫ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﺴﻠﺴﻠﺔ ‪) D12 p t‬‬

‫‪230‬‬
 ‫}‪{3‬‬
‫‪z t =. 0936 + 0. 97 z‬‬ ‫‪t -1‬‬ ‫‪+ e t + -0. 6 e t - 1‬‬
‫) ‪( 0. 1‬‬ ‫) ‪( 0. 068‬‬
‫‪RSS = 662 S 2 = 5. 62 T = 119‬‬
‫ﺍﳌﻄﻠﻮﺏ‪:‬‬
‫‪ -1-‬ﺗﻌﻘﹼﺐ ﺧﻄﻮﺍﺕ ﻫﺬﺍ ﺍﻟﺒﺎﺣﺚ ﻣﻊ ﺫﻛﺮ ﻛﻞ ﻣﺎ ﻟﻪ ﻭﻣﺎ ﻋﻠﻴﻪ ﺑﺎﺳﺘﺨﺪﺍﻡ‬
‫ﻛﻞ ﺍﻟﻮﺳﺎﺋﻞ ﺍﻟﻨﻈﺮﻳﺔ ﻭﺍﻹﺣﺼﺎﺋﻴﺔ ﺍﻟﱵ ﺗﻌﺮﻓﻬﺎ ﻣﻊ ﻛﺘﺎﺑﺔ ﻋﻼﻗﺔ ﻛﻞ ﻭﺳﻴﻠﺔ‬
‫ﺍﺳﺘﺨﺪﻣﻬﺎ‪.‬‬
‫‪ -2-‬ﺑﻐﻴﺔ ﺍﻻﺳﺘﻔﺎﺩﺓ ﺃﻛﺜﺮ‪ ،‬ﻧﻘﺘﺮﺡ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻹﺿﺎﻓﻴﺔ ﺍﻟﺘﺎﻟﻴﺔ ﺍﳌﺴﺘﻨﺘﺠﺔ ﻣﻦ‬
‫ﺍﻟﺴﻠﺴﻠﺔ ‪ ، D12 p t = sp‬ﻓﻬﻞ ﺗﺆﺛﺮ ﺷﻴﺌﺎﹰ ﻋﻠﻰ ﺍﻟﻘﺮﺍﺭ ﻓﻴﻤﺎ ﺇﺫﺍ ﺩﺧﻠﺖ ﳎﺎﻝ‬
‫ﺍﳌﻔﺎﺿﻠﺔ ؟‪.‬‬

‫}‪{4‬‬
‫‪z t = 0. 95 z t - 1 - 0. 65 e t -1 + 0. 26 e t - 5‬‬
‫) ‪( 0. 05‬‬ ‫)‪0. 07‬‬ ‫)‪( 0. 7‬‬
‫‪R 2 = 0. 664 RSS = 612. 46 s = 2. 3 T = 119‬‬

‫}‪{5‬‬
‫‪z t = 1. 86 + 0. 3z t -3 + 0. 3z t -6 + 0. 43e t -1 + 0. 32e t -2 + 0. 23e t -5‬‬
‫) ‪( 0. 58) ( 0. 09) ( 0. 8 ) ( 0. 08) ( 0. 1‬‬ ‫)‪( 0. 09‬‬
‫‪R 2 = 0. 64 RSS= 618. 1 s= 2. 39‬‬ ‫‪T= 114‬‬

‫}‪{6‬‬
‫‪z t = 0. 27z t-3 + 0. 69z t-5 + 0. 32e t-1 + 0. 37et -2 - 0. 55e t-5‬‬
‫)‪( 0. 8‬‬ ‫)‪( 0.08) ( 0.07) ( 0. 07) ( 0.08‬‬
‫‪R2 = 0. 69 RSS= 595.9 s= 2. 32‬‬ ‫‪T=115‬‬

‫‪231‬‬
 160
[G1]

140

P
120

100

80
78 79 80 81 82 83 84 85 86 87 88 89

232
Included observations: 132 F1 (P)
Autocorrelation Partial Correlation AC PAC Q-Stat Prob
.|*** | .|*** | 1 0.457 0.457 28.221 0.000
.|** | .|. | 2 0.226 0.022 35.179 0.000
.|* | .|* | 3 0.170 0.075 39.156 0.000
.|** | .|* | 4 0.210 0.131 45.260 0.000
.|** | .|* | 5 0.221 0.087 52.049 0.000
.|* | .|. | 6 0.183 0.037 56.772 0.000
.|* | .|* | 7 0.196 0.091 62.227 0.000
.|* | .|. | 8 0.171 0.024 66.399 0.000
.|* | .|. | 9 0.114 -0.024 68.266 0.000
.|* | .|* | 10 0.146 0.075 71.341 0.000
.|*** | .|** | 11 0.351 0.290 89.320 0.000
.|****** | .|****** | 12 0.810 0.751 185.92 0.000
.|*** | ****|. | 13 0.334 -0.517 202.50 0.000
.|* | *|. | 14 0.129 -0.141 205.00 0.000
.|* | *|. | 15 0.077 -0.098 205.89 0.000
.|* | *|. | 16 0.103 -0.137 207.51 0.000
.|* | *|. | 17 0.103 -0.107 209.16 0.000
.|* | .|. | 18 0.072 0.010 209.96 0.000
.|* | .|. | 19 0.091 0.034 211.27 0.000
.|* | .|. | 20 0.072 0.022 212.10 0.000
.|. | .|. | 21 0.019 0.039 212.15 0.000
.|. | .|. | 22 0.046 0.035 212.50 0.000
.|** | .|. | 23 0.227 -0.043 220.89 0.000
.|***** | .|. | 24 0.635 0.050 287.02 0.000
.|** | .|. | 25 0.218 -0.046 294.85 0.000
.|. | .|. | 26 0.043 -0.012 295.16 0.000
.|. | .|. | 27 -0.005 -0.028 295.16 0.000
.|. | .|. | 28 0.018 0.012 295.22 0.000
.|. | .|. | 29 0.017 0.005 295.27 0.000
.|. | .|. | 30 -0.012 -0.023 295.29 0.000
.|. | .|. | 31 0.007 -0.031 295.30 0.000
.|. | .|. | 32 -0.014 -0.044 295.33 0.000
.|. | .|. | 33 -0.054 0.022 295.85 0.000
.|. | .|. | 34 -0.034 -0.029 296.05 0.000
.|* | .|. | 35 0.131 0.006 299.17 0.000
.|**** | .|. | 36 0.491 -0.001 343.65 0.000
.|* | *|. | 37 0.108 -0.085 345.83 0.000
.|. | .|. | 38 -0.042 -0.018 346.15 0.000
*|. | .|. | 39 -0.085 -0.018 347.52 0.000
*|. | .|. | 40 -0.058 0.004 348.17 0.000
.|. | .|. | 41 -0.056 0.012 348.79 0.000
*|. | .|. | 42 -0.082 0.013 350.10 0.000
*|. | .|. | 43 -0.059 0.027 350.78 0.000
*|. | .|. | 44 -0.078 -0.013 352.02 0.000
*|. | .|. | 45 -0.111 -0.023 354.54 0.000
*|. | .|. | 46 -0.093 -0.007 356.30 0.000
.|. | .|. | 47 0.056 -0.032 356.95 0.000
.|*** | .|. | 48 0.376 -0.013 386.74 0.000
.|. | .|. | 49 0.032 0.010 386.96 0.000
*|. | .|. | 50 -0.098 -0.013 389.03 0.000

233
cluded observations: 131
F2 (Dp)
Autocorrelation Partial Correlation AC PAC Q-Stat Prob
**|. | **|. | 1 -0.256 -0.256 8.7953 0.003
*|. | **|. | 2 -0.159 -0.241 12.216 0.002
*|. | **|. | 3 -0.087 -0.226 13.241 0.004
.|. | *|. | 4 0.026 -0.137 13.331 0.010
.|. | *|. | 5 0.040 -0.073 13.551 0.019
.|. | *|. | 6 -0.038 -0.103 13.757 0.032
.|. | .|. | 7 0.043 -0.015 14.013 0.051
.|. | .|. | 8 0.032 0.028 14.163 0.078
*|. | *|. | 9 -0.088 -0.076 15.268 0.084
*|. | **|. | 10 -0.172 -0.256 19.541 0.034
**|. | *****|. | 11 -0.264 -0.604 29.661 0.002
.|*******| .|****** | 12 0.865 0.717 139.24 0.000
**|. | .|. | 13 -0.221 0.032 146.46 0.000
*|. | .|. | 14 -0.134 0.013 149.12 0.000
*|. | .|. | 15 -0.069 0.038 149.84 0.000
.|. | .|. | 16 0.026 0.018 149.94 0.000
.|. | .|. | 17 0.029 -0.039 150.07 0.000
.|. | *|. | 18 -0.048 -0.069 150.43 0.000
.|. | .|. | 19 0.050 -0.026 150.82 0.000
.|. | *|. | 20 0.025 -0.115 150.92 0.000
*|. | .|. | 21 -0.068 -0.012 151.64 0.000
*|. | .|. | 22 -0.155 0.043 155.51 0.000
**|. | .|* | 23 -0.232 0.129 164.15 0.000
.|****** | .|. | 24 0.751 0.053 256.01 0.000
**|. | .|. | 25 -0.199 0.010 262.49 0.000
*|. | .|. | 26 -0.108 0.039 264.43 0.000
*|. | *|. | 27 -0.069 -0.066 265.22 0.000
.|. | .|. | 28 0.028 -0.039 265.35 0.000
.|. | .|. | 29 0.024 -0.036 265.45 0.000
.|. | .|. | 30 -0.046 0.005 265.82 0.000
.|. | .|. | 31 0.048 -0.045 266.22 0.000
.|. | .|. | 32 0.018 0.011 266.27 0.000
.|. | .|. | 33 -0.057 0.019 266.84 0.000
*|. | .|. | 34 -0.140 0.024 270.35 0.000
**|. | .|. | 35 -0.203 0.013 277.86 0.000
.|***** | .|. | 36 0.667 0.059 359.44 0.000
*|. | .|. | 37 -0.179 0.008 365.39 0.000
*|. | .|. | 38 -0.090 -0.037 366.90 0.000
*|. | .|. | 39 -0.064 0.007 367.67 0.000
.|. | .|. | 40 0.021 -0.039 367.75 0.000
.|. | .|. | 41 0.026 0.006 367.89 0.000
.|. | .|. | 42 -0.043 -0.009 368.25 0.000
.|. | .|. | 43 0.044 0.000 368.64 0.000
.|. | .|. | 44 0.014 0.012 368.67 0.000
.|. | .|. | 45 -0.050 0.003 369.17 0.000
*|. | .|. | 46 -0.127 0.018 372.50 0.000
*|. | .|. | 47 -0.181 0.002 379.31 0.000
.|**** | .|. | 48 0.584 -0.032 450.95 0.000
*|. | .|. | 49 -0.153 0.006 455.90 0.000
*|. | .|. | 50 -0.078 -0.015 457.20 0.000

234
ncluded observations: 120 F3 (D12p)
Autocorrelation Partial Correlation AC PAC Q-Stat Prob
.|****** | .|****** | 1 0.748 0.748 68.851 0.000
.|***** | .|*** | 2 0.707 0.335 130.90 0.000
.|***** | .|* | 3 0.661 0.149 185.54 0.000
.|***** | .|* | 4 0.636 0.117 236.65 0.000
.|***** | .|** | 5 0.667 0.225 293.22 0.000
.|***** | .|. | 6 0.622 0.022 342.93 0.000
.|***** | .|. | 7 0.617 0.064 392.31 0.000
.|**** | *|. | 8 0.549 -0.095 431.65 0.000
.|**** | .|. | 9 0.514 -0.049 466.54 0.000
.|**** | *|. | 10 0.478 -0.068 496.92 0.000
.|*** | *|. | 11 0.430 -0.087 521.76 0.000
.|*** | *|. | 12 0.365 -0.173 539.83 0.000
.|*** | .|* | 13 0.374 0.078 558.95 0.000
.|*** | .|. | 14 0.336 -0.021 574.54 0.000
.|*** | .|* | 15 0.332 0.070 589.88 0.000
.|** | .|. | 16 0.313 0.056 603.71 0.000
.|** | .|. | 17 0.252 -0.034 612.73 0.000
.|** | .|. | 18 0.234 -0.002 620.61 0.000
.|** | .|* | 19 0.230 0.100 628.27 0.000
.|** | .|. | 20 0.219 -0.013 635.28 0.000
.|* | *|. | 21 0.176 -0.088 639.87 0.000
.|* | *|. | 22 0.139 -0.084 642.76 0.000
.|* | .|. | 23 0.120 -0.040 644.94 0.000
.|* | *|. | 24 0.080 -0.110 645.91 0.000
.|* | .|. | 25 0.078 0.009 646.86 0.000
.|* | .|. | 26 0.066 -0.003 647.54 0.000
.|. | .|. | 27 0.020 -0.046 647.60 0.000
.|. | .|. | 28 0.017 0.061 647.65 0.000
.|. | .|. | 29 -0.019 -0.020 647.71 0.000
*|. | *|. | 30 -0.065 -0.098 648.40 0.000
*|. | .|. | 31 -0.081 0.014 649.47 0.000
*|. | .|* | 32 -0.071 0.089 650.32 0.000
*|. | .|. | 33 -0.060 0.064 650.93 0.000
*|. | .|. | 34 -0.068 0.036 651.70 0.000
.|. | .|* | 35 -0.057 0.073 652.25 0.000
*|. | .|. | 36 -0.086 -0.038 653.54 0.000
*|. | .|* | 37 -0.066 0.119 654.32 0.000
*|. | .|. | 38 -0.058 0.031 654.92 0.000
.|. | .|. | 39 -0.053 -0.042 655.42 0.000
.|. | .|. | 40 -0.057 -0.039 656.01 0.000
*|. | *|. | 41 -0.079 -0.108 657.18 0.000
.|. | .|. | 42 -0.039 0.003 657.46 0.000
.|. | .|. | 43 -0.037 0.020 657.73 0.000
.|. | *|. | 44 -0.041 -0.066 658.06 0.000
.|. | .|. | 45 -0.035 0.010 658.30 0.000
.|. | .|. | 46 -0.052 0.020 658.84 0.000
.|. | .|. | 47 -0.042 0.030 659.19 0.000
*|. | **|. | 48 -0.090 -0.190 660.86 0.000
*|. | .|* | 49 -0.072 0.068 661.92 0.000
*|. | .|. | 50 -0.082 0.004 663.34 0.000

235
Included observations: 119 F4 (Dd12p)
Autocorrelation Partial Correlation AC PAC Q-Stat Prob
***|. | ***|. | 1 -0.412 -0.412 20.688 0.000
.|. | **|. | 2 0.006 -0.196 20.693 0.000
.|. | *|. | 3 -0.046 -0.155 20.952 0.000
*|. | **|. | 4 -0.118 -0.262 22.704 0.000
.|* | *|. | 5 0.132 -0.078 24.907 0.000
.|. | *|. | 6 -0.042 -0.073 25.135 0.000
.|* | .|* | 7 0.113 0.075 26.766 0.000
.|. | .|. | 8 -0.049 0.046 27.081 0.001
.|. | .|. | 9 -0.020 0.039 27.135 0.001
.|. | .|* | 10 0.027 0.069 27.232 0.002
.|. | .|* | 11 0.032 0.142 27.371 0.004
*|. | *|. | 12 -0.155 -0.118 30.600 0.002
.|* | .|. | 13 0.108 -0.018 32.192 0.002
*|. | *|. | 14 -0.064 -0.086 32.750 0.003
.|. | .|. | 15 0.052 -0.041 33.118 0.005
.|* | .|. | 16 0.085 0.058 34.120 0.005
*|. | .|. | 17 -0.087 0.020 35.181 0.006
.|. | *|. | 18 -0.038 -0.082 35.387 0.008
.|. | .|. | 19 0.022 0.032 35.455 0.012
.|. | .|* | 20 0.045 0.077 35.754 0.016
.|. | .|* | 21 0.012 0.067 35.775 0.023
.|. | .|. | 22 -0.044 -0.007 36.068 0.030
.|. | .|* | 23 0.060 0.079 36.601 0.036
*|. | *|. | 24 -0.101 -0.071 38.138 0.034
.|. | *|. | 25 0.004 -0.075 38.141 0.045
.|* | .|. | 26 0.078 -0.026 39.075 0.048
*|. | *|. | 27 -0.098 -0.120 40.564 0.045
.|* | .|. | 28 0.082 -0.022 41.633 0.047
.|. | .|* | 29 0.014 0.067 41.666 0.060
*|. | .|. | 30 -0.068 -0.049 42.421 0.066
.|. | *|. | 31 -0.037 -0.095 42.647 0.079
.|. | *|. | 32 -0.010 -0.067 42.663 0.099
.|. | .|. | 33 0.041 -0.026 42.943 0.115
.|. | *|. | 34 -0.031 -0.075 43.103 0.136
.|* | .|. | 35 0.090 0.053 44.492 0.131
*|. | *|. | 36 -0.104 -0.131 46.365 0.116
.|. | *|. | 37 0.010 -0.061 46.384 0.139
.|. | .|. | 38 0.018 0.024 46.443 0.164
.|. | .|. | 39 0.019 0.019 46.510 0.191
.|. | .|. | 40 0.019 0.061 46.578 0.220
*|. | .|. | 41 -0.107 -0.036 48.702 0.191
.|* | .|. | 42 0.071 -0.045 49.639 0.195
.|. | .|. | 43 0.013 0.049 49.671 0.225
.|. | .|. | 44 -0.016 -0.008 49.718 0.256
.|. | .|. | 45 0.034 -0.011 49.937 0.284
.|. | .|. | 46 -0.050 -0.014 50.427 0.303
.|* | .|** | 47 0.094 0.202 52.192 0.279
*|. | *|. | 48 -0.138 -0.100 56.065 0.198
.|* | .|. | 49 0.085 -0.008 57.568 0.188
.|. | .|. | 50 -0.036 -0.008 57.840 0.208

236
 ‫‪:8‬‬ ‫ﲤﺮﻳﻦ‬
‫‪ -‬ﺍ‪ /‬ﺑﲔ ﺃﳘﻴﺔ ﺩﺭﺍﺳﺔ ﻣﺸﻜﻞ ﺍﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ؟ ﻭﺍﺫﻛﺮ ﺍﳌﺸﺎﻛﻞ‬
‫ﺍﻟﻨﺎﲨﺔ ﻋﻦ ﻋﺪﻡ ﺍﺳﺘﻘﺮﺍﺭ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ؟‬
‫ﺏ‪ /‬ﺑﲔ ﺟﱪﻳﺎﹰ ﺃﻥ ﺗﺒﺎﻳﻦ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻏﲑ ﺍﳌﺴﺘﻘﺮﺓ ﻳﻨﻤﻮ ﻣﻊ ﺍﻟﺰﻣﻦ ﰲ‬
‫ﺍﻟﻨﻤﻮﺫﺝ ﺍﻟﺘﺎﱄ‪:‬‬
‫‪yt = yt -1 + e t‬‬
‫ﺝ‪ /‬ﻋﺮﻑ ﺍﻟﺘﻘﺪﻳﺮ ﺍﳌﺰﻳﻒ ﻭﺑﲔ ﻣﻈﺎﻫﺮﻩ ﺍﻟﻘﻴﺎﺳﻴﺔ‪.‬‬

‫‪:9‬‬ ‫ﲤﺮﻳﻦ‬
‫)‪(38‬‬ ‫‪-‬ﺑﺎﻟﺮﺟﻮﻉ ﺇﱃ ﻧﻔﺲ ﻣﻌﻄﻴﺎﺕ ﺍﻟﻜﺘﻠﺔ ﺍﻟﻨﻘﺪﻳﺔ ‪ M2‬ﺍﻟﻮﺍﺭﺩﺓ ﰲ ﺍﳉﺪﻭﻝ‬
‫ﺻﻔﺤﺔ )‪ ،(208‬ﺍﳌﻄﻠﻮﺏ ﲢﺪﻳﺪ ﻣﺪﻯ ﺍﺳﺘﻘﺮﺍﺭﻳﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺰﻣﻨﻴﺔ ﻫﺬﻩ ﺑﺎﻋﺘﻤﺎﺩ‬
‫ﺍﺧﺘﺒﺎﺭ ‪ Φ‬ﻟﺪﻳﻜﻲ ﻓﻮﻟﺮ )ﻓﺮﺿﻴﺔ ﻣﺸﺘﺮﻛﺔ(‪ .‬ﻓﺈﺫﺍ ﻛﺎﻧﺖ ﻏﲑ ﻣﺴﺘﻘﺮﺓ ﺍﳌﻄﻠﻮﺏ‬
‫ﺍﺧﺘﻴﺎﺭ ﺍﻟﻄﺮﻳﻘﺔ ﺍﻷﻧﺴﺐ ﻟﺘﺤﻮﻳﻠﻬﺎ ﺇﱃ ﺍﻟﺸﻜﻞ ﺍﳌﺴﺘﻘﺮ‪.‬‬

‫‪237‬‬
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
 ‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫@‬ ‫@‬
‫اﳉﺪاول اﻻﺣﺼﺎﺋﻴﺔ‬
‫@ @‬
‫@ @‬
‫@ @‬

‫‪239‬‬
 @ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @
@ @

242
 ‫ﺟﺪﻭﻝ ‪ :1‬ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ ﻻﺧﺘﺒﺎﺭ ‪،R‬‬
‫‪Ru‬‬ ‫‪Rl‬‬ ‫‪m‬‬ ‫‪Ru‬‬ ‫‪Rl‬‬ ‫‪m‬‬
‫‪19‬‬ ‫‪8‬‬ ‫‪13‬‬ ‫‪10‬‬ ‫‪2‬‬ ‫‪5‬‬
‫‪20‬‬ ‫‪9‬‬ ‫‪14‬‬ ‫‪11‬‬ ‫‪3‬‬ ‫‪6‬‬
‫‪22‬‬ ‫‪10‬‬ ‫‪15‬‬ ‫‪13‬‬ ‫‪3‬‬ ‫‪7‬‬
‫‪23‬‬ ‫‪11‬‬ ‫‪16‬‬ ‫‪14‬‬ ‫‪4‬‬ ‫‪8‬‬
‫‪25‬‬ ‫‪11‬‬ ‫‪17‬‬ ‫‪15‬‬ ‫‪5‬‬ ‫‪9‬‬
‫‪26‬‬ ‫‪12‬‬ ‫‪18‬‬ ‫‪16‬‬ ‫‪6‬‬ ‫‪10‬‬
‫‪27‬‬ ‫‪13‬‬ ‫‪19‬‬ ‫‪17‬‬ ‫‪7‬‬ ‫‪11‬‬
‫‪28‬‬ ‫‪14‬‬ ‫‪20‬‬ ‫‪17‬‬ ‫‪7‬‬ ‫‪12‬‬

‫ﺟﺪﻭﻝ ‪ :2‬ﺍﻟﻘﻴﻢ ﺍﳊﺮﺟﺔ ﳌﻌﺎﻣﻞ ﺍﻻﺭﺗﺒﺎﻁ ﻟﺴﺒﲑﻣﺎﻥ‬

‫‪0.001‬‬ ‫‪0.005‬‬ ‫‪0.01‬‬ ‫‪0.025‬‬ ‫‪0.05‬‬ ‫‪0.1‬‬ ‫‪1 a‬‬
‫‪2‬‬ ‫‪n‬‬
‫‪0.002‬‬ ‫‪0.01‬‬ ‫‪0.02‬‬ ‫‪0.05‬‬ ‫‪0.1‬‬ ‫‪0.2‬‬ ‫‪a‬‬
‫‪-‬‬ ‫‪-‬‬ ‫‪-‬‬ ‫‪-‬‬ ‫‪0.8‬‬ ‫‪0.8‬‬ ‫‪4‬‬
‫‪-‬‬ ‫‪-‬‬ ‫‪0.9‬‬ ‫‪0.9‬‬ ‫‪0.8‬‬ ‫‪0.7‬‬ ‫‪5‬‬
‫‪-‬‬ ‫‪0.9429‬‬ ‫‪0.8857‬‬ ‫‪0.8286‬‬ ‫‪0.7714‬‬ ‫‪0.6‬‬ ‫‪6‬‬
‫‪0.9643‬‬ ‫‪0.8929‬‬ ‫‪0.8571‬‬ ‫‪0.745‬‬ ‫‪0.6789‬‬ ‫‪0.5357‬‬ ‫‪7‬‬
‫‪0.9286‬‬ ‫‪0.8571‬‬ ‫‪0.8095‬‬ ‫‪0.7143‬‬ ‫‪0.619‬‬ ‫‪0.5‬‬ ‫‪8‬‬
‫‪0.9‬‬ ‫‪0.8167‬‬ ‫‪0.7667‬‬ ‫‪0.6833‬‬ ‫‪0.5833‬‬ ‫‪0.4667‬‬ ‫‪9‬‬
‫‪0.8667‬‬ ‫‪0.7818‬‬ ‫‪0.7333‬‬ ‫‪0.6364‬‬ ‫‪0.5515‬‬ ‫‪0.4424‬‬ ‫‪10‬‬
‫‪0.8364‬‬ ‫‪0.7545‬‬ ‫‪0.07‬‬ ‫‪0.609‬‬ ‫‪0.5273‬‬ ‫‪0.4182‬‬ ‫‪11‬‬
‫‪0.8182‬‬ ‫‪0.7273‬‬ ‫‪0.6713‬‬ ‫‪0.5804‬‬ ‫‪0.4965‬‬ ‫‪0.3986‬‬ ‫‪12‬‬
‫‪0.7912‬‬ ‫‪0.6978‬‬ ‫‪0.6429‬‬ ‫‪0.5549‬‬ ‫‪0.478‬‬ ‫‪0.3791‬‬ ‫‪13‬‬
‫‪0.767‬‬ ‫‪0.6747‬‬ ‫‪0.622‬‬ ‫‪0.5341‬‬ ‫‪0.4593‬‬ ‫‪0.3626‬‬ ‫‪14‬‬
‫‪0.7464‬‬ ‫‪0.6536‬‬ ‫‪0.6‬‬ ‫‪0.5179‬‬ ‫‪0.4429‬‬ ‫‪0.35‬‬ ‫‪15‬‬
‫‪0.7265‬‬ ‫‪0.6324‬‬ ‫‪0.5824‬‬ ‫‪0.5‬‬ ‫‪0.4265‬‬ ‫‪0.3382‬‬ ‫‪16‬‬
‫‪0.7083‬‬ ‫‪0.6152‬‬ ‫‪0.5637‬‬ ‫‪0.4853‬‬ ‫‪0.4118‬‬ ‫‪0.326‬‬ ‫‪17‬‬
‫‪0.6904‬‬ ‫‪0.5975‬‬ ‫‪0.5480‬‬ ‫‪0.4716‬‬ ‫‪0.3994‬‬ ‫‪0.3148‬‬ ‫‪18‬‬
‫‪0.6737‬‬ ‫‪0.5825‬‬ ‫‪0.5333‬‬ ‫‪0.4779‬‬ ‫‪0.3895‬‬ ‫‪0.3070‬‬ ‫‪19‬‬
‫‪0.6586‬‬ ‫‪0.5684‬‬ ‫‪0.5203‬‬ ‫‪0.4451‬‬ ‫‪0.3789‬‬ ‫‪0.2977‬‬ ‫‪20‬‬
‫‪0.6455‬‬ ‫‪0.5545‬‬ ‫‪0.5078‬‬ ‫‪0.4351‬‬ ‫‪0.3688‬‬ ‫‪0.299‬‬ ‫‪21‬‬
‫‪0.6318‬‬ ‫‪0.5426‬‬ ‫‪0.4963‬‬ ‫‪0.4241‬‬ ‫‪0.3597‬‬ ‫‪0.2829‬‬ ‫‪22‬‬
‫‪0.6186‬‬ ‫‪0.5306‬‬ ‫‪0.4852‬‬ ‫‪0.415‬‬ ‫‪0.3518‬‬ ‫‪0.2767‬‬ ‫‪23‬‬
‫‪0.6070‬‬ ‫‪0.52‬‬ ‫‪0.4748‬‬ ‫‪0.4061‬‬ ‫‪0.3435‬‬ ‫‪0.2704‬‬ ‫‪24‬‬
‫‪0.5962‬‬ ‫‪0.51‬‬ ‫‪0.4654‬‬ ‫‪0.3977‬‬ ‫‪0.3362‬‬ ‫‪0.2646‬‬ ‫‪25‬‬
‫‪0.5856‬‬ ‫‪0.5002‬‬ ‫‪0.4564‬‬ ‫‪0.3894‬‬ ‫‪0.3299‬‬ ‫‪0.2588‬‬ ‫‪26‬‬
‫‪0.5757‬‬ ‫‪0.4915‬‬ ‫‪0.4481‬‬ ‫‪0.3822‬‬ ‫‪0.3236‬‬ ‫‪0.254‬‬ ‫‪27‬‬
‫‪0.568‬‬ ‫‪0.4828‬‬ ‫‪0.4401‬‬ ‫‪0.3749‬‬ ‫‪0.3175‬‬ ‫‪0.249‬‬ ‫‪28‬‬
‫‪0.5567‬‬ ‫‪0.4744‬‬ ‫‪0.432‬‬ ‫‪0.3685‬‬ ‫‪0.3113‬‬ ‫‪0.2443‬‬ ‫‪29‬‬
‫‪0.5479‬‬ ‫‪0.4665‬‬ ‫‪0.4251‬‬ ‫‪0.362‬‬ ‫‪0.3059‬‬ ‫‪0.24‬‬ ‫‪30‬‬

‫‪243‬‬
 ‫ﺟﺪﻭﻝ ‪ :3‬ﺍﻟﺘﻮﺯﻳﻊ ﺍﻟﻄﺒﻴﻌﻲ ﺍﳌﻌﻴﺎﺭﻱ‬
‫‪.09‬‬ ‫‪.08‬‬ ‫‪.07‬‬ ‫‪.06‬‬ ‫‪.05‬‬ ‫‪.04‬‬ ‫‪.02‬‬ ‫‪.01‬‬ ‫‪.00‬‬ ‫‪z‬‬
‫‪.0359‬‬ ‫‪.0319‬‬ ‫‪.0279‬‬ ‫‪.0239‬‬ ‫‪.0199‬‬ ‫‪.0160‬‬ ‫‪.0080‬‬ ‫‪.0040‬‬ ‫‪.0000‬‬ ‫‪0.0‬‬
‫‪.0753‬‬ ‫‪.0714‬‬ ‫‪.0675‬‬ ‫‪.0636‬‬ ‫‪.0596‬‬ ‫‪.0557‬‬ ‫‪.0487‬‬ ‫‪.0438‬‬ ‫‪.0398‬‬ ‫‪0.1‬‬
‫‪.1141‬‬ ‫‪.1103‬‬ ‫‪.1064‬‬ ‫‪.1026‬‬ ‫‪.0987‬‬ ‫‪.0948‬‬ ‫‪.0871‬‬ ‫‪.0832‬‬ ‫‪.0793‬‬ ‫‪0.2‬‬
‫‪.1517‬‬ ‫‪.1480‬‬ ‫‪.1443‬‬ ‫‪.1406‬‬ ‫‪.1368‬‬ ‫‪.1331‬‬ ‫‪.1255‬‬ ‫‪.1217‬‬ ‫‪.1179‬‬ ‫‪0.3‬‬
‫‪.1879‬‬ ‫‪.1844‬‬ ‫‪.1808‬‬ ‫‪.1772‬‬ ‫‪.1736‬‬ ‫‪.1700‬‬ ‫‪.1628‬‬ ‫‪.1591‬‬ ‫‪.1554‬‬ ‫‪0.4‬‬
‫‪.2224‬‬ ‫‪.2190‬‬ ‫‪.2157‬‬ ‫‪.2123‬‬ ‫‪.2088‬‬ ‫‪.2054‬‬ ‫‪.1985‬‬ ‫‪.1950‬‬ ‫‪.1915‬‬ ‫‪0.5‬‬
‫‪.2549‬‬ ‫‪.2517‬‬ ‫‪.2486‬‬ ‫‪.2454‬‬ ‫‪.2422‬‬ ‫‪.2389‬‬ ‫‪.2324‬‬ ‫‪.2291‬‬ ‫‪.2257‬‬ ‫‪0.6‬‬
‫‪.2852‬‬ ‫‪.2823‬‬ ‫‪.2794‬‬ ‫‪.2764‬‬ ‫‪.2734‬‬ ‫‪.2703‬‬ ‫‪.2642‬‬ ‫‪.2611‬‬ ‫‪.2580‬‬ ‫‪0.7‬‬
‫‪.3133‬‬ ‫‪.3106‬‬ ‫‪.3078‬‬ ‫‪.3051‬‬ ‫‪.3023‬‬ ‫‪.2995‬‬ ‫‪.2939‬‬ ‫‪.2910‬‬ ‫‪.2881‬‬ ‫‪0.8‬‬
‫‪.3389‬‬ ‫‪.3365‬‬ ‫‪.3340‬‬ ‫‪.3315‬‬ ‫‪.3289‬‬ ‫‪.3264‬‬ ‫‪.3213‬‬ ‫‪.3186‬‬ ‫‪.3159‬‬ ‫‪0.9‬‬
‫‪.3621‬‬ ‫‪.3599‬‬ ‫‪.3577‬‬ ‫‪.3554‬‬ ‫‪.3531‬‬ ‫‪.3508‬‬ ‫‪.3461‬‬ ‫‪.3438‬‬ ‫‪.3413‬‬ ‫‪1.0‬‬
‫‪.3830‬‬ ‫‪.3910‬‬ ‫‪.3790‬‬ ‫‪.3770‬‬ ‫‪.3749‬‬ ‫‪.3729‬‬ ‫‪.3686‬‬ ‫‪.3665‬‬ ‫‪.3643‬‬ ‫‪1.1‬‬
‫‪.4015‬‬ ‫‪.3997‬‬ ‫‪.3980‬‬ ‫‪.3962‬‬ ‫‪.3944‬‬ ‫‪.3925‬‬ ‫‪.3888‬‬ ‫‪.3869‬‬ ‫‪.3849‬‬ ‫‪1.2‬‬
‫‪.4177‬‬ ‫‪.4162‬‬ ‫‪.4147‬‬ ‫‪.4131‬‬ ‫‪.4115‬‬ ‫‪.4099‬‬ ‫‪.4066‬‬ ‫‪.4049‬‬ ‫‪.4032‬‬ ‫‪1.3‬‬
‫‪.4319‬‬ ‫‪.4306‬‬ ‫‪.4292‬‬ ‫‪.4279‬‬ ‫‪.4265‬‬ ‫‪.4251‬‬ ‫‪.4222‬‬ ‫‪.4207‬‬ ‫‪.4192‬‬ ‫‪1.4‬‬
‫‪.4441‬‬ ‫‪.4429‬‬ ‫‪.4418‬‬ ‫‪.4406‬‬ ‫‪.4394‬‬ ‫‪.4382‬‬ ‫‪.4357‬‬ ‫‪.4345‬‬ ‫‪.4332‬‬ ‫‪1.5‬‬
‫‪.4545‬‬ ‫‪.4535‬‬ ‫‪.4525‬‬ ‫‪.4515‬‬ ‫‪.4505‬‬ ‫‪.4495‬‬ ‫‪.4474‬‬ ‫‪.4463‬‬ ‫‪.4452‬‬ ‫‪1.6‬‬
‫‪.4633‬‬ ‫‪.4625‬‬ ‫‪.4616‬‬ ‫‪.4608‬‬ ‫‪.4599‬‬ ‫‪.4591‬‬ ‫‪.4573‬‬ ‫‪.4564‬‬ ‫‪.4554‬‬ ‫‪1.7‬‬
‫‪.4706‬‬ ‫‪.4699‬‬ ‫‪.4693‬‬ ‫‪.4686‬‬ ‫‪.4678‬‬ ‫‪.4671‬‬ ‫‪.4656‬‬ ‫‪.4649‬‬ ‫‪.4641‬‬ ‫‪1.8‬‬
‫‪.4767‬‬ ‫‪.4761‬‬ ‫‪.4756‬‬ ‫‪.4750‬‬ ‫‪.4744‬‬ ‫‪.4738‬‬ ‫‪.4726‬‬ ‫‪.4719‬‬ ‫‪.4713‬‬ ‫‪1.9‬‬
‫‪.4817‬‬ ‫‪.4812‬‬ ‫‪.4808‬‬ ‫‪.4803‬‬ ‫‪.4798‬‬ ‫‪.4793‬‬ ‫‪.4783‬‬ ‫‪.4778‬‬ ‫‪.4772‬‬ ‫‪2.0‬‬
‫‪.4857‬‬ ‫‪.4854‬‬ ‫‪.4850‬‬ ‫‪.4846‬‬ ‫‪.4842‬‬ ‫‪.4838‬‬ ‫‪.4830‬‬ ‫‪.4829‬‬ ‫‪.4821‬‬ ‫‪2.1‬‬
‫‪.4890‬‬ ‫‪.4887‬‬ ‫‪.4884‬‬ ‫‪.4881‬‬ ‫‪.4878‬‬ ‫‪.4875‬‬ ‫‪.4868‬‬ ‫‪.4864‬‬ ‫‪.4861‬‬ ‫‪2.2‬‬
‫‪.4916‬‬ ‫‪.4913‬‬ ‫‪.4911‬‬ ‫‪.4909‬‬ ‫‪.4906‬‬ ‫‪.4904‬‬ ‫‪.4898‬‬ ‫‪.4896‬‬ ‫‪.4893‬‬ ‫‪2.3‬‬
‫‪.4936‬‬ ‫‪.4934‬‬ ‫‪.4932‬‬ ‫‪.4931‬‬ ‫‪.4929‬‬ ‫‪.4927‬‬ ‫‪.4922‬‬ ‫‪.4920‬‬ ‫‪.4918‬‬ ‫‪2.4‬‬
‫‪.4952‬‬ ‫‪.4951‬‬ ‫‪.4949‬‬ ‫‪.4948‬‬ ‫‪.4946‬‬ ‫‪.4945‬‬ ‫‪.4941‬‬ ‫‪.4940‬‬ ‫‪.4938‬‬ ‫‪2.5‬‬
‫‪.4964‬‬ ‫‪.4963‬‬ ‫‪.4962‬‬ ‫‪.4961‬‬ ‫‪.4960‬‬ ‫‪.4959‬‬ ‫‪.4956‬‬ ‫‪.4955‬‬ ‫‪.4953‬‬ ‫‪2.6‬‬
‫‪.4974‬‬ ‫‪.4973‬‬ ‫‪.4972‬‬ ‫‪.4971‬‬ ‫‪.4970‬‬ ‫‪.4969‬‬ ‫‪.4967‬‬ ‫‪.4966‬‬ ‫‪.4965‬‬ ‫‪2.7‬‬
‫‪.4981‬‬ ‫‪.4980‬‬ ‫‪.4979‬‬ ‫‪4979‬‬ ‫‪.4978‬‬ ‫‪.4977‬‬ ‫‪.4976‬‬ ‫‪.4975‬‬ ‫‪.4974‬‬ ‫‪2.8‬‬
‫‪.4986‬‬ ‫‪.4986‬‬ ‫‪.4985‬‬ ‫‪.4985‬‬ ‫‪.4984‬‬ ‫‪.4984‬‬ ‫‪.4982‬‬ ‫‪.4982‬‬ ‫‪.4981‬‬ ‫‪2.9‬‬
‫‪.4990‬‬ ‫‪.4990‬‬ ‫‪.4989‬‬ ‫‪.4989‬‬ ‫‪.4989‬‬ ‫‪.4988‬‬ ‫‪.4987‬‬ ‫‪.4987‬‬ ‫‪.4987‬‬ ‫‪3.0‬‬

‫‪244‬‬
 Student's t ‫ ﺗﻮﺯﻳﻊ‬:4 ‫ﺟﺪﻭﻝ‬
.005 .01 .025 .05 .10 P
63.657 31.821 12.706 6.314 3.078 1
9.925 6.965 4.303 2.920 1.886 2
5.841 4.541 3.182 2.353 1.638 3
4.604 3.747 2.776 2.132 1.533 4
4.032 3.365 2.571 2.015 1.476 5
3.707 3.143 2.447 1.943 1.440 6
3.499 2.998 2.365 1.895 1.415 7
3.355 2.896 2.306 1.860 1.397 8
3.250 2.821 2.262 1.833 1.383 9
3.169 2.764 2.228 1.812 1.372 10
3.106 2.718 2.201 1.796 1.363 11
3.055 2.681 2.179 1.782 1.356 12
3.012 2.650 2.160 1.771 1.350 13
2.977 2.624 2.145 1.761 1.345 14
2.947 2.602 2.131 1.753 1.341 15
2.921 2.583 2.120 1.746 1.337 16
2.898 2.567 2.110 1.740 1.333 17
2.878 2.552 2.101 1.734 1.330 18
2.861 2.539 2.093 1.729 1.328 19
2.845 2.528 2.086 1.725 1.325 20
2.831 2.518 2.080 1.721 1.323 21
2.819 2.508 2.074 1.717 1.321 22
2.807 2.500 2.069 1.714 1.319 23
2.797 2.492 2.064 1.711 1.318 24
2.787 2.485 2.060 1.708 1.316 25
2.779 2.479 2.056 1.706 1.315 26
2.771 2.473 2.052 1.703 1.314 27
2.763 2.467 2.048 1.701 1.313 28
2.756 2.462 2.045 1.699 1.311 29
2.750 2.457 2.042 1.697 1.310 30

245
 ‫‪ :5‬ﺗﻮﺯﻳﻊ@‪Chi2‬‬ ‫ﺟﺪﻭﻝ‬
‫‪0.05‬‬ ‫‪0.10‬‬ ‫‪0.50‬‬ ‫‪0.90‬‬ ‫‪0.95‬‬ ‫‪p=0.99‬‬ ‫‪DF‬‬
‫‪3.841‬‬ ‫‪2.706‬‬ ‫‪0.455‬‬ ‫‪0.0158‬‬ ‫‪0.00393‬‬ ‫‪0.000157‬‬ ‫‪1‬‬
‫‪5.991‬‬ ‫‪4.605‬‬ ‫‪1.386‬‬ ‫‪0.211‬‬ ‫‪0.103‬‬ ‫‪0.0201‬‬ ‫‪2‬‬
‫‪7.815‬‬ ‫‪6.251‬‬ ‫‪2.366‬‬ ‫‪0.584‬‬ ‫‪0.352‬‬ ‫‪0.115‬‬ ‫‪3‬‬
‫‪9.488‬‬ ‫‪7.779‬‬ ‫‪3.357‬‬ ‫‪1.064‬‬ ‫‪0.711‬‬ ‫‪0.297‬‬ ‫‪4‬‬
‫‪11.070‬‬ ‫‪9.236‬‬ ‫‪4.351‬‬ ‫‪1.610‬‬ ‫‪1.145‬‬ ‫‪0.554‬‬ ‫‪5‬‬
‫‪3.070‬‬ ‫‪10.645‬‬ ‫‪5.348‬‬ ‫‪2.204‬‬ ‫‪1.635‬‬ ‫‪0.872‬‬ ‫‪6‬‬
‫‪14.067‬‬ ‫‪12.017‬‬ ‫‪6.346‬‬ ‫‪2.833‬‬ ‫‪2.167‬‬ ‫‪1.239‬‬ ‫‪7‬‬
‫‪15.507‬‬ ‫‪13.362‬‬ ‫‪7.344‬‬ ‫‪3.490‬‬ ‫‪2.733‬‬ ‫‪1.646‬‬ ‫‪8‬‬
‫‪16.919‬‬ ‫‪14.684‬‬ ‫‪8.343‬‬ ‫‪4.168‬‬ ‫‪3.325‬‬ ‫‪2.088‬‬ ‫‪9‬‬
‫‪18.307‬‬ ‫‪15.987‬‬ ‫‪9.342‬‬ ‫‪4.865‬‬ ‫‪3.940‬‬ ‫‪2.558‬‬ ‫‪10‬‬
‫‪19.675‬‬ ‫‪17.275‬‬ ‫‪10.341‬‬ ‫‪5.578‬‬ ‫‪4.575‬‬ ‫‪3.053‬‬ ‫‪11‬‬
‫‪21.026‬‬ ‫‪18.549‬‬ ‫‪11.340‬‬ ‫‪6.304‬‬ ‫‪5.226‬‬ ‫‪3.571‬‬ ‫‪12‬‬
‫‪22.362‬‬ ‫‪19.812‬‬ ‫‪12.340‬‬ ‫‪7.042‬‬ ‫‪5.892‬‬ ‫‪4.107‬‬ ‫‪13‬‬
‫‪21.064‬‬ ‫‪18.151‬‬ ‫‪10.821‬‬ ‫‪6.571‬‬ ‫‪5.368‬‬ ‫‪4.660‬‬ ‫‪14‬‬
‫‪24.996‬‬ ‫‪22.307‬‬ ‫‪14.339‬‬ ‫‪8.547‬‬ ‫‪7.261‬‬ ‫‪5.229‬‬ ‫‪15‬‬
‫‪26.296‬‬ ‫‪23.542‬‬ ‫‪15.338‬‬ ‫‪9.312‬‬ ‫‪7.962‬‬ ‫‪5.812‬‬ ‫‪16‬‬
‫‪27.587‬‬ ‫‪24.769‬‬ ‫‪16.338‬‬ ‫‪10.085‬‬ ‫‪8.672‬‬ ‫‪6.408‬‬ ‫‪17‬‬
‫‪28.869‬‬ ‫‪25.989‬‬ ‫‪17.338‬‬ ‫‪10.865‬‬ ‫‪9.390‬‬ ‫‪7.015‬‬ ‫‪18‬‬
‫‪30.144‬‬ ‫‪27.204‬‬ ‫‪18.338‬‬ ‫‪11.651‬‬ ‫‪10.117‬‬ ‫‪7.633‬‬ ‫‪19‬‬
‫‪31.410‬‬ ‫‪28.412‬‬ ‫‪19.337‬‬ ‫‪12.443‬‬ ‫‪10.851‬‬ ‫‪8.260‬‬ ‫‪20‬‬
‫‪32.671‬‬ ‫‪29.615‬‬ ‫‪20.337‬‬ ‫‪13.240‬‬ ‫‪11..591‬‬ ‫‪8.897‬‬ ‫‪21‬‬
‫‪33.924‬‬ ‫‪30.813‬‬ ‫‪21.337‬‬ ‫‪14.041‬‬ ‫‪12.338‬‬ ‫‪9.542‬‬ ‫‪22‬‬
‫‪35.172‬‬ ‫‪32.007‬‬ ‫‪22.337‬‬ ‫‪14.848‬‬ ‫‪13.091‬‬ ‫‪10.196‬‬ ‫‪23‬‬
‫‪36.415‬‬ ‫‪33.196‬‬ ‫‪23.337‬‬ ‫‪15.659‬‬ ‫‪13.848‬‬ ‫‪10.856‬‬ ‫‪24‬‬
‫‪37.652‬‬ ‫‪34.382‬‬ ‫‪24.337‬‬ ‫‪16.473‬‬ ‫‪14.611‬‬ ‫‪11.524‬‬ ‫‪25‬‬
‫‪38.885‬‬ ‫‪35.563‬‬ ‫‪25.336‬‬ ‫‪17.292‬‬ ‫‪15.379‬‬ ‫‪12.198‬‬ ‫‪26‬‬
‫‪40.113‬‬ ‫‪36.741‬‬ ‫‪26.336‬‬ ‫‪18.114‬‬ ‫‪16.151‬‬ ‫‪12.879‬‬ ‫‪27‬‬
‫‪41.337‬‬ ‫‪37.916‬‬ ‫‪27.336‬‬ ‫‪18.939‬‬ ‫‪16.928‬‬ ‫‪13.565‬‬ ‫‪28‬‬
‫‪42.557‬‬ ‫‪39.087‬‬ ‫‪28.336‬‬ ‫‪19.768‬‬ ‫‪17.708‬‬ ‫‪14.256‬‬ ‫‪29‬‬
‫‪43.773‬‬ ‫‪40.256‬‬ ‫‪29.336‬‬ ‫‪20.599‬‬ ‫‪18.493‬‬ ‫‪14.953‬‬ ‫‪30‬‬

‫‪246‬‬
 ‫‪F‬‬ ‫ﺟﺪﻭﻝ ‪ :5‬ﺗﻮﺯﻳﻊ ﻓﻴﺸﺮ‬
‫ﺩﺭﺟﺎﺕ ﺣﺮﻳﺔ ﺍﻟﺒﺴﻂ‬ ‫ﺩﺭﺟﺎﺕ‬
‫‪11‬‬ ‫‪10‬‬ ‫‪9‬‬ ‫‪8‬‬ ‫‪7‬‬ ‫‪6‬‬ ‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫ﺣﺮﻳﺔ ﺍﳌﻘﺎﻡ‬
‫‪243‬‬ ‫‪242‬‬ ‫‪241‬‬ ‫‪239‬‬ ‫‪237‬‬ ‫‪234‬‬ ‫‪230‬‬ ‫‪225‬‬ ‫‪216‬‬ ‫‪200‬‬ ‫‪161‬‬ ‫‪5%‬‬
‫‪1‬‬
‫‪6082‬‬ ‫‪6056‬‬ ‫‪6022‬‬ ‫‪5981‬‬ ‫‪5928‬‬ ‫‪5859‬‬ ‫‪5764‬‬ ‫‪5625‬‬ ‫‪5403‬‬ ‫‪4999‬‬ ‫‪4052‬‬ ‫‪1%‬‬
‫‪19.40‬‬ ‫‪19.39‬‬ ‫‪19.38‬‬ ‫‪19.37‬‬ ‫‪19.36‬‬ ‫‪19.33‬‬ ‫‪19.30‬‬ ‫‪19.25‬‬ ‫‪19.16‬‬ ‫‪19.00‬‬ ‫‪18.51‬‬ ‫‪5%‬‬
‫‪2‬‬
‫‪99.41‬‬ ‫‪99.40‬‬ ‫‪99.38‬‬ ‫‪99.36‬‬ ‫‪99.34‬‬ ‫‪99.33‬‬ ‫‪99.30‬‬ ‫‪99.25‬‬ ‫‪99.17‬‬ ‫‪99.01‬‬ ‫‪98.49‬‬ ‫‪1%‬‬
‫‪8.76‬‬ ‫‪8.78‬‬ ‫‪8.81‬‬ ‫‪8.84‬‬ ‫‪8.88‬‬ ‫‪8.94‬‬ ‫‪9.01‬‬ ‫‪9.12‬‬ ‫‪9.28‬‬ ‫‪9.55‬‬ ‫‪10.13‬‬ ‫‪5%‬‬
‫‪3‬‬
‫‪27.13‬‬ ‫‪27.23‬‬ ‫‪27.34‬‬ ‫‪27.49‬‬ ‫‪27.67‬‬ ‫‪27.91‬‬ ‫‪28.24‬‬ ‫‪28.71‬‬ ‫‪29.46‬‬ ‫‪30.81‬‬ ‫‪34.12‬‬ ‫‪1%‬‬
‫‪5.93‬‬ ‫‪5.96‬‬ ‫‪6.00‬‬ ‫‪6.04‬‬ ‫‪6.09‬‬ ‫‪6.16‬‬ ‫‪6.26‬‬ ‫‪6.39‬‬ ‫‪6.59‬‬ ‫‪6.94‬‬ ‫‪7.71‬‬ ‫‪5%‬‬
‫‪4‬‬
‫‪14.47‬‬ ‫‪14.54‬‬ ‫‪14.66‬‬ ‫‪14.80‬‬ ‫‪14.98‬‬ ‫‪15.21‬‬ ‫‪15.52‬‬ ‫‪15.98‬‬ ‫‪16.69‬‬ ‫‪18.00‬‬ ‫‪21.30‬‬ ‫‪1%‬‬
‫‪4.70‬‬ ‫‪4.75‬‬ ‫‪4.78‬‬ ‫‪4.82‬‬ ‫‪4.88‬‬ ‫‪4.95‬‬ ‫‪5.05‬‬ ‫‪5.19‬‬ ‫‪5.41‬‬ ‫‪5.79‬‬ ‫‪6.61‬‬ ‫‪5%‬‬
‫‪5‬‬
‫‪9.96‬‬ ‫‪10.05‬‬ ‫‪10.15‬‬ ‫‪10.27‬‬ ‫‪10.45‬‬ ‫‪10.67‬‬ ‫‪10.97‬‬ ‫‪11.39‬‬ ‫‪12.06‬‬ ‫‪13.27‬‬ ‫‪16.26‬‬ ‫‪1%‬‬
‫‪4.03‬‬ ‫‪4.06‬‬ ‫‪4.10‬‬ ‫‪4.15‬‬ ‫‪4.21‬‬ ‫‪4.28‬‬ ‫‪4.39‬‬ ‫‪4.53‬‬ ‫‪4.76‬‬ ‫‪5.14‬‬ ‫‪5.99‬‬ ‫‪5%‬‬
‫‪6‬‬
‫‪7.79‬‬ ‫‪7.87‬‬ ‫‪7.98‬‬ ‫‪8.10‬‬ ‫‪8.26‬‬ ‫‪8.47‬‬ ‫‪8.75‬‬ ‫‪9.15‬‬ ‫‪9.78‬‬ ‫‪10.92‬‬ ‫‪13.74‬‬ ‫‪1%‬‬
‫‪3.60‬‬ ‫‪3.63‬‬ ‫‪3.68‬‬ ‫‪3.73‬‬ ‫‪3.79‬‬ ‫‪3.87‬‬ ‫‪3.97‬‬ ‫‪4.12‬‬ ‫‪4.35‬‬ ‫‪4.74‬‬ ‫‪5.59‬‬ ‫‪5%‬‬
‫‪7‬‬
‫‪6.54‬‬ ‫‪6.62‬‬ ‫‪6.71‬‬ ‫‪6.84‬‬ ‫‪7.00‬‬ ‫‪7.19‬‬ ‫‪7.46‬‬ ‫‪7.85‬‬ ‫‪8.45‬‬ ‫‪9.55‬‬ ‫‪12.25‬‬ ‫‪1%‬‬
‫‪3.31‬‬ ‫‪3.34‬‬ ‫‪3.39‬‬ ‫‪3.44‬‬ ‫‪3.50‬‬ ‫‪3.58‬‬ ‫‪3.69‬‬ ‫‪3.84‬‬ ‫‪4.07‬‬ ‫‪4.46‬‬ ‫‪5.32‬‬ ‫‪5%‬‬
‫‪8‬‬
‫‪5.74‬‬ ‫‪5.82‬‬ ‫‪5.91‬‬ ‫‪6.03‬‬ ‫‪6.19‬‬ ‫‪6.37‬‬ ‫‪6.63‬‬ ‫‪7.01‬‬ ‫‪7.59‬‬ ‫‪8.65‬‬ ‫‪11.26‬‬ ‫‪1%‬‬
‫‪3.10‬‬ ‫‪3.13‬‬ ‫‪3.18‬‬ ‫‪3.23‬‬ ‫‪3.29‬‬ ‫‪3.37‬‬ ‫‪3.48‬‬ ‫‪3.63‬‬ ‫‪3.86‬‬ ‫‪4.26‬‬ ‫‪5.12‬‬ ‫‪5%‬‬
‫‪9‬‬
‫‪5.18‬‬ ‫‪5.26‬‬ ‫‪5.35‬‬ ‫‪5.47‬‬ ‫‪5.62‬‬ ‫‪5.80‬‬ ‫‪6.06‬‬ ‫‪6.42‬‬ ‫‪6.99‬‬ ‫‪8.02‬‬ ‫‪10.56‬‬ ‫‪1%‬‬
‫‪2.94‬‬ ‫‪2.97‬‬ ‫‪3.02‬‬ ‫‪3.07‬‬ ‫‪3.14‬‬ ‫‪3.22‬‬ ‫‪3.33‬‬ ‫‪3.48‬‬ ‫‪3.71‬‬ ‫‪4.10‬‬ ‫‪4.96‬‬ ‫‪5%‬‬
‫‪10‬‬
‫‪4.78‬‬ ‫‪4.85‬‬ ‫‪4.95‬‬ ‫‪5.06‬‬ ‫‪5.21‬‬ ‫‪5.39‬‬ ‫‪5.64‬‬ ‫‪5.99‬‬ ‫‪6.55‬‬ ‫‪7.56‬‬ ‫‪10.04‬‬ ‫‪1%‬‬
‫‪2.82‬‬ ‫‪2.96‬‬ ‫‪2.90‬‬ ‫‪2.95‬‬ ‫‪3.01‬‬ ‫‪3.09‬‬ ‫‪3.20‬‬ ‫‪3.36‬‬ ‫‪3.59‬‬ ‫‪3.98‬‬ ‫‪4.84‬‬ ‫‪5%‬‬
‫‪11‬‬
‫‪4.46‬‬ ‫‪4.54‬‬ ‫‪4.63‬‬ ‫‪4.74‬‬ ‫‪4.88‬‬ ‫‪5.07‬‬ ‫‪5.32‬‬ ‫‪5.67‬‬ ‫‪6.22‬‬ ‫‪7.20‬‬ ‫‪9.65‬‬ ‫‪1%‬‬
‫‪2.72‬‬ ‫‪2.76‬‬ ‫‪2.80‬‬ ‫‪2.85‬‬ ‫‪2.92‬‬ ‫‪3.00‬‬ ‫‪3.11‬‬ ‫‪3.26‬‬ ‫‪3.49‬‬ ‫‪3.88‬‬ ‫‪4.75‬‬ ‫‪5%‬‬
‫‪12‬‬
‫‪4.22‬‬ ‫‪4.30‬‬ ‫‪4.39‬‬ ‫‪4.50‬‬ ‫‪4.65‬‬ ‫‪4.82‬‬ ‫‪5.06‬‬ ‫‪5.41‬‬ ‫‪5.95‬‬ ‫‪6.93‬‬ ‫‪9.33‬‬ ‫‪1%‬‬
‫‪2.63‬‬ ‫‪2.67‬‬ ‫‪2.72‬‬ ‫‪2.77‬‬ ‫‪2.84‬‬ ‫‪2.92‬‬ ‫‪3.02‬‬ ‫‪3.18‬‬ ‫‪3.41‬‬ ‫‪3.80‬‬ ‫‪4.67‬‬ ‫‪5%‬‬
‫‪13‬‬
‫‪4.02‬‬ ‫‪4.10‬‬ ‫‪4.19‬‬ ‫‪4.30‬‬ ‫‪4.44‬‬ ‫‪4.62‬‬ ‫‪4.86‬‬ ‫‪5.20‬‬ ‫‪5.74‬‬ ‫‪6.70‬‬ ‫‪9.07‬‬ ‫‪1%‬‬
‫‪2.56‬‬ ‫‪2.60‬‬ ‫‪2.65‬‬ ‫‪2.70‬‬ ‫‪2.77‬‬ ‫‪2.85‬‬ ‫‪2.96‬‬ ‫‪3.11‬‬ ‫‪3.34‬‬ ‫‪3.74‬‬ ‫‪4.60‬‬ ‫‪5%‬‬
‫‪14‬‬
‫‪3.86‬‬ ‫‪3.94‬‬ ‫‪4.03‬‬ ‫‪4.14‬‬ ‫‪4.28‬‬ ‫‪4.46‬‬ ‫‪4.69‬‬ ‫‪5.03‬‬ ‫‪5.56‬‬ ‫‪6.51‬‬ ‫‪8.86‬‬ ‫‪1%‬‬
‫‪2.51‬‬ ‫‪2.55‬‬ ‫‪2.59‬‬ ‫‪2.64‬‬ ‫‪2.70‬‬ ‫‪2.79‬‬ ‫‪2.90‬‬ ‫‪3.06‬‬ ‫‪3.29‬‬ ‫‪3.68‬‬ ‫‪4.54‬‬ ‫‪5%‬‬
‫‪15‬‬
‫‪3.73‬‬ ‫‪3.80‬‬ ‫‪3.89‬‬ ‫‪4.00‬‬ ‫‪4.14‬‬ ‫‪4.32‬‬ ‫‪4.56‬‬ ‫‪4.89‬‬ ‫‪5.42‬‬ ‫‪6.36‬‬ ‫‪8.68‬‬ ‫‪1%‬‬
‫‪2.45‬‬ ‫‪2.49‬‬ ‫‪2.54‬‬ ‫‪2.59‬‬ ‫‪2.66‬‬ ‫‪2.74‬‬ ‫‪2.85‬‬ ‫‪3.01‬‬ ‫‪3.24‬‬ ‫‪3.63‬‬ ‫‪4.49‬‬ ‫‪5%‬‬
‫‪16‬‬
‫‪3.61‬‬ ‫‪3.69‬‬ ‫‪3.78‬‬ ‫‪3.89‬‬ ‫‪4.03‬‬ ‫‪4.20‬‬ ‫‪4.44‬‬ ‫‪4.77‬‬ ‫‪5.29‬‬ ‫‪6.23‬‬ ‫‪8.53‬‬ ‫‪1%‬‬
‫‪2.41‬‬ ‫‪2.45‬‬ ‫‪2.50‬‬ ‫‪2.55‬‬ ‫‪2.62‬‬ ‫‪2.70‬‬ ‫‪2.81‬‬ ‫‪2.96‬‬ ‫‪3.20‬‬ ‫‪3.59‬‬ ‫‪4.45‬‬ ‫‪5%‬‬
‫‪17‬‬
‫‪3.52‬‬ ‫‪3.59‬‬ ‫‪3.68‬‬ ‫‪3.79‬‬ ‫‪3.93‬‬ ‫‪4.10‬‬ ‫‪4.34‬‬ ‫‪4.67‬‬ ‫‪5.18‬‬ ‫‪6.11‬‬ ‫‪8.40‬‬ ‫‪1%‬‬
‫‪2.37‬‬ ‫‪2.41‬‬ ‫‪2.46‬‬ ‫‪2.51‬‬ ‫‪2.58‬‬ ‫‪2.66‬‬ ‫‪2.77‬‬ ‫‪2.93‬‬ ‫‪3.16‬‬ ‫‪3.55‬‬ ‫‪4.41‬‬ ‫‪5%‬‬
‫‪18‬‬
‫‪3.44‬‬ ‫‪3.51‬‬ ‫‪3.60‬‬ ‫‪3.71‬‬ ‫‪3.85‬‬ ‫‪4.01‬‬ ‫‪4.25‬‬ ‫‪4.58‬‬ ‫‪5.09‬‬ ‫‪6.01‬‬ ‫‪8.28‬‬ ‫‪1%‬‬
‫‪2.34‬‬ ‫‪2.38‬‬ ‫‪2.43‬‬ ‫‪2.48‬‬ ‫‪2.55‬‬ ‫‪2.63‬‬ ‫‪2.74‬‬ ‫‪2.90‬‬ ‫‪3.13‬‬ ‫‪3.52‬‬ ‫‪4.38‬‬ ‫‪5%‬‬
‫‪19‬‬
‫‪3.36‬‬ ‫‪3.43‬‬ ‫‪3.52‬‬ ‫‪3.63‬‬ ‫‪3.77‬‬ ‫‪3.94‬‬ ‫‪4.17‬‬ ‫‪4.50‬‬ ‫‪5.01‬‬ ‫‪3.93‬‬ ‫‪8.18‬‬ ‫‪1%‬‬
‫‪2.31‬‬ ‫‪2.35‬‬ ‫‪2.40‬‬ ‫‪2.45‬‬ ‫‪2.52‬‬ ‫‪2.60‬‬ ‫‪2.71‬‬ ‫‪2.87‬‬ ‫‪3.10‬‬ ‫‪3.49‬‬ ‫‪4.35‬‬ ‫‪5%‬‬
‫‪20‬‬
‫‪3.30‬‬ ‫‪3.37‬‬ ‫‪3.45‬‬ ‫‪3.56‬‬ ‫‪3.71‬‬ ‫‪3.87‬‬ ‫‪4.10‬‬ ‫‪4.43‬‬ ‫‪4.94‬‬ ‫‪5.85‬‬ ‫‪8.10‬‬ ‫‪1%‬‬

‫‪247‬‬
 F ‫ ﺗﻮﺯﻳﻊ ﻓﻴﺸﺮ‬:5 ‫ﺟﺪﻭﻝ‬
¡éj¤a @ÚÌäy@p b uâÖ ÚÌäy@p b uâÖ
11 10 9 8 7 6 5 4 3 2 1 ‚ b‘Ωa@
2.28 2.32 2.37 2.42 2.49 2.57 2.68 2.84 3.07 3.47 4.32 5%
21
3.24 3.31 3.40 3.51 3.65 3.81 4.04 4.37 4.87 5.78 8.02 1%
2.26 2.30 2.35 2.40 2.47 2.55 2.66 2.82 3.05 3.44 4.30 5%
22
3.18 3.26 3.35 3.45 3.59 3.76 3.99 4.31 4.82 5.72 7.94 1%
2.24 2.28 2.32 2.38 2.45 2.53 2.64 2.80 3.03 3.42 4.28 5%
23
3.14 3.21 3.30 3.41 3.54 3.71 3.94 4.26 4.76 5.66 7.88 1%
2.22 2.26 2.30 2.36 2.43 2.51 2.62 2.78 3.01 3.40 4.26 5%
24
3.09 3.17 3.25 3.36 3.50 3.67 3.90 4.22 4.72 5.61 7.82 1%
2.20 2.24 2.28 2.34 2.41 2.49 2.60 2.76 2.99 3.38 4.24 5%
25
3.05 3.13 3.21 3.32 3.46 3.63 3.86 4.18 4.68 5.57 7.77 1%
2.18 2.22 2.27 2.32 2.39 2.47 2.59 2.74 2.89 3.37 4.22 5%
26
3.02 3.09 3.17 3.29 3.42 3.59 3.82 4.14 4.64 5.53 7.72 1%
2.16 2.20 2.25 2.30 2.37 2.46 2.57 2.73 2.96 3.35 4.21 5%
27
2.98 3.06 3.14 3.26 3.39 3.56 3.79 4.11 4.60 5.49 7.68 1%
2.15 2.19 3.24 2.29 2.36 2.44 2.56 2.71 2.95 3.34 4.20 5%
28
2.95 3.03 3.11 3.23 3.36 3.53 3.76 4.07 4.57 5.45 7.64 1%
2.14 2.18 2.22 2.28 2.35 2.43 2.54 2.70 2.93 3.33 4.18 5%
29
2.92 3.00 3.08 3.20 3.33 3.50 3.73 4.04 4.54 5.52 7.60 1%
2.12 2.16 2.21 2.27 2.34 2.42 2.53 2.69 2.92 3.32 4.17 5%
30
2.90 2.98 3.06 3.17 3.30 3.47 3.70 4.02 4.51 5.39 7.56 1%
2.10 2.14 2.19 2.25 2.32 2.40 2.51 2.67 2.90 3.30 4.15 5%
32
2.86 2.94 3.01 3.12 3.25 3.42 3.66 3.97 4.46 5.34 7.50 1%
2.08 2.12 2.17 2.23 2.30 2.38 2.49 2.65 2.88 3.28 4.13 5%
34
2.82 2.89 2.97 3.08 3.21 3.38 3.61 3.93 4.42 5.29 7.44 1%
2.06 2.10 2.15 2.21 2.28 2.36 2.48 2.63 2.86 3.26 4.11 5%
36
2.78 2.86 2.94 3.04 3.18 3.35 3.58 3.89 4.38 5.25 7.39 1%
2.05 2.09 2.14 2.19 2.26 2.35 2.46 2.62 2.85 3.25 4.10 5%
38
2.75 2.82 2.91 3.02 3.15 3.32 3.54 3.86 4.34 5.21 7.35 1%
2.04 2.07 2.12 2.18 2.25 2.34 2.45 2.61 2.84 3.23 4.08 5%
40
2.73 2.80 2.88 2.99 3.12 3.29 3.51 3.83 4.31 5.18 7.31 1%
2.02 2.06 2.11 2.17 2.24 2.32 2.44 2.59 2.83 3.22 4.07 5%
42
2.70 2.77 2.86 2.96 3.10 3.26 3.49 3.80 4.29 5.15 7.27 1%
2.01 2.05 2.10 2.16 2.23 2.31 2.43 2.58 2.82 3.21 4.06 5%
44
2.68 2.75 2.84 2.94 3.07 3.24 3.46 3.78 4.26 5.12 7.24 1%
2.00 2.04 2.09 2.14 2.22 2.30 2.42 2.57 2.81 3.20 4.05 5%
46
2.66 2.73 2.82 2.92 3.05 3.22 3.44 3.76 4.24 5.10 7.21 1%
1.99 2.03 2.08 2.14 2.21 2.30 2.41 2.56 2.80 3.19 4.04 5%
48
2.64 2.71 2.80 2.90 3.04 3.20 3.42 3.74 4.22 5.08 7.19 1%
1.98 2.02 2.07 2.13 2.20 2.29 2.40 2.56 2.79 3.18 4.03 5%
50
2.62 2.70 2.78 2.88 3.02 3.18 3.41 3.72 4.20 5.06 7.17 1%

248
 ‫ﺍﳌﺮﺍﺟﻊ‬
‫ﺍﳌﺮﺍﺟﻊ ﺍﻟﻌﺮﺑﻴﺔ‪:‬‬ ‫‪-1‬‬

‫]‪ -[1‬ﺗﻮﻣﻲ ﺹ‪" ،(1999) .‬ﻣﺪﺧﻞ ﻟﻨﻈﺮﻳﺔ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ"‪ ،‬ﺝ‪ ،1‬ﺝ‪،2‬‬
‫ﺩ‪.‬ﻡ‪.‬ﺝ‪ .‬ﺍﳉﺰﺍﺋﺮ‪.‬‬
‫]‪ -[2‬ﺟﻼﻃﻮ ﺝ‪" ،(2005) .‬ﺍﻹﺣﺼﺎﺀ"‪ ،‬ﺩ‪.‬ﻡ‪.‬ﺝ‪ .‬ﺍﳉﺰﺍﺋﺮ‪.‬‬
‫]‪ -[3‬ﺣﺸﻤﺎﻥ ﻡ‪ .‬ﻭﻣﺴﻠﻢ ﻉ‪ ،(2008) .‬ﺍﲡﺎﻫﺎﺕ ﺍﻟﻨﻤﻮ ﺍﻻﻗﺘﺼﺎﺩﻱ ﰲ ﺍﳉﺰﺍﺋﺮ‬
‫ﺧﻼﻝ ﺍﻟﻔﺘﺮﺓ ‪ ،2004-1990‬ﳎﻠﺔ ﻋﻠﻮﻡ ﺍﻻﻗﺘﺼﺎﺩ ﻭﺍﻟﺘﺴﻴﲑ ﻭﺍﻟﺘﺠﺎﺭﺓ‪ ،‬ﺍﻟﻌﺪﺩ‬
‫‪ ،17‬ﺍ‪‬ﻠﺪ‪.1‬‬
‫]‪ -[4‬ﺣﺸﻤﺎﻥ ﻡ‪" ،(2002) .‬ﳕﺎﺫﺝ ﻭﺗﻘﻨﻴﺎﺕ ﺍﻟﺘﻨﺒﺆ ﺍﻟﻘﺼﲑ ﺍﳌﺪﻯ" ﺩ‪.‬ﻡ‪.‬ﺝ‪.‬‬
‫ﺍﳉﺰﺍﺋﺮ‪.‬‬
‫]‪ -[5‬ﻉ‪.‬ﻉ‪ .‬ﺷﺮﻳﻒ )‪" ،(1982‬ﻣﻘﺪﻣﺔ ﰲ ﺍﻟﻘﻴﺎﺱ ﺍﻻﻗﺘﺼﺎﺩﻱ"‪ ،‬ﺩ‪.‬ﻡ‪.‬ﺝ‪ .‬ﺍﳉﺰﺍﺋﺮ‪.‬‬
‫]‪ -[6‬ﺍﻟﺘﻘﺮﻳﺮ ﺍﻻﻗﺘﺼﺎﺩﻱ ﺍﻟﻌﺮﰊ ﺍﳌﻮﺣﺪ‪ ،‬ﺳﺒﺘﻤﱪ )‪.(2005‬‬

‫‪249‬‬
 ‫ ﺍﳌﺮﺍﺟﻊ ﺍﻷﺟﻨﺒﻴﺔ‬- 2
[7]- Ansion G. (1990), "Les méthodes de Prévision en Economie", ARMAND
COLIN.
[8]- Antoniadis A., S.Berruyer, R.Carmona (1992) "Régression non Linéaire et
Application", ECONOMICA.
[9]- Arestis P., G. Hadjimatheou (1982), “Introducing Macroeconomic
Modeling, an Econometric study of the United kingdom”, MAC MILLAN
PRESS LTD.
[10]- Bailly P. (1993), "Exercices Corriges de Statistique Descriptive", OPU
Alger.
[11]- Bourbonnais R., Terrasa M. (2008); "Analyses des séries temporelles,
applications à l’économie et à la gestion", 2eme édition, DUNOD, Paris.
[12]- Bowers David (1982), "Statistics for economists”, MAC MILLAN PRESS
LTD.
[13]- Box G.E.P., G.M.Jenkins (1971)," Time Series Analysis - forecasting and
control”, HOLDEN DAY.
[14]- Caron N., P. Tassi (1991), "Problèmes Résolus de Statistique
Mathématique”, ECONOMICA.
[15]- Christ C. F. (1966), “Econometric Models and Methods”, WILEY.
[16]- David M., J. C. Michaud (1989), "La Prévision, Approche Empirique d'une
Méthode Statistique“, Masson.
[17]- Dicky D., A. and W. A.Fuller (1981), "Likelihood Ratio Statistics for
Autoregressive Time- Series with a Unit Root," Econometrica, 49, 1057-
1072.
[18]- Droesbeke J. J. , B . Fichet, P. Tassi (1989), "Séries Chronologiques”,
Paris.
[19]- Farnum N. R., Laverne W. Stanton (1989), "Quantitative Forecasting
Methods", PWS - Kent Publishing Company Boston.
[20]- Gourieroux C., A. Monfort, Vol.2 (1989), "Statistique et Modèles
Econométriques", ECONOMICA.
[21]- Gourieroux C., A. Monfort (1990), "Séries Temporelles et Modèles
Dynamiques”, ECONOMICA, Paris.
[22]- Granger C.W.J, P. Newbold (1974), "Spurious regression in econometrics
",Journal of econometrics,2,1045- 66.
[23]- Granger C.W.J, P. Newbold (1986), "Forecasting Economic Time Series ",
2nd Edition, ACADEMIC PRESS INC.
[24]- Harvey A. C. (1981)," Time Series Models" PHILIP ALLAN.
[25]- Harvey A. C. (1990), "The Econometric Analysis of Time Series”, PHILIP
ALLAN.
[26]- Hoel P. G. (1984), "Introduction to Mathematical Statistics", JOHN
WILEY&SONS.

250
[27]- Holden K. (1995),"Vector Auto regression Modeling and forecasting",
Journal of Forecasting, Vol. 14, 159-166.
[28]- Hoskins R J., R. J. Goult, J.A.Milner, M.J. Pratt (1974), Computational
Methods in Linear Algebra , STANLEY THORNES.
[29]- Johnson R.A., D.W.Wichern (1992), "Applied Multivariate statistical
Analysis", PRENTICE HALL, New Jersey.
[30]- Joutz F.L, G.S.Maddala, R.P.Trost, (1995), "An Integrated Vector Auto
regression and Error Correction Model for Forecasting Electricity
consumption and prices" , Journal of Forecasting ,Vol. 14 , 287-310.
[31]- Judge G. G. and all (1982), "Introduction to the Theory and Practice of
Econometrics”, WILEY.
[32]- Kaufman, J.L. Groboillot (1967), "La Prévision Economique à Court
Terme", DUNOD.
[33]- Kmenta J., J. B. Ramsey (1980), “Evaluation of Econometric Models”,
ACADEMIC PRESS.
[34]- Lewandowski R. (1979), "La Prévision à Court Terme", DUNOD.
[35]- Maddala G.S. (1979), “Econometrics”, INTERNATIONAL STUDENT
EDITION.
[36]- Maddala G.S.,In-Moo Kim (2000), “Unit roots, co-integration and
structural change”, Cambridge University Press.
[37]- Mandell B.J. (1972), “statistics for management”, DANDARY
PUBLISHING COMPANY, INC, USA.
[38]- Montgomery D. C., Lynwood A. Johson (1976), "Forecasting and Time -
Series Analysis “, MC GRAWHILL BOOK Company.
[39]- Monfort A. (1992), "Quelques Développements Récent Des Méthodes
Macro-économétriques", Actualité Economique. Revue d'Analyse
Economique, Vol 68.
[40]- Osborn D. R. (1982), "Testing for Causality in Multivariate ARMA
Models”, University of Manchester, July 1982.
[41]- Osborn D.R. (1994), “A Critical Evaluation of the Use of sequential
Probabilities in predicting cyclical Turning points”, Discussion paper
ES249, Manchester University.
[42]- Pichot A. (1988),"Comptabilité Nationale et Modèles Economiques“, PUF.
[43]- Pindyck R. S., D. L. Rubinfeld (1981), “Econometric Models and
Economic Forecasts”, INTERNATIONAL STUDENT EDITION.
[44]- Sims C. (1980),"Macroeconomics and Reality", Econometrica, 48, pp.1-48.
[45]- Stewart J. (1991), “Econometrics”, PHILIP ALLAN.
[46]- Tenenbaus M. (1994), "Méthodes Statistiques en Gestion”, DUNOD.
[47]- Pena D. (1995),"Forecasting growth With Time Series models”, Journal of
Forecasting, Vol. 14, 97 - 105.
[48]- Webb R.H. (1995)," Forecasts of inflation from VAR Models “Journal of
Forecasting, Vol. 14, 267 - 285.

251
[49]- Willard C.J. (1992), "Aspects Actuels de L'Economie Quantitative",
L'HARMATAN.
[50]- Wine R. L.,(1976); ”Beginning statistics”, WINTHROP PUBLISHERS,
INC.
[51]- Wheelwright S. C., S. Makridakis (1983),"Méthodes de Prévision pour la
Gestion", Les Editions d'Organisation.
[52]- Wonnacott T, Wonnacott RJ. (1990), “Econometrics”, Introductory
statistics for business and economics”, 4th edition, WILEY.

252