‫)‪2010 (17‬‬ ‫מא‬ ‫א‬ ‫א‬ ‫א‬

‫א‬ ‫‪−‬א‬ ‫א‬ ‫א‬ ‫א‬

‫]‪[ 128 −115‬‬

‫ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻟﻠﺘﺨﻠﺹ ﻤﻥ ﺘﻌﺩﺩ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺨﻁﻴﺔ‬

‫***‬ ‫**‬ ‫*‬

‫ﺼﺎﻟﺢ ﻤﺅﻴﺩ ﺸﺎﻜﺭ‬ ‫ﺴﻴﻑ ﺍﻟﺩﻴﻥ ﻀﻴﺎﺀ ﺍﻟﺩﻴﻥ‬ ‫ﺼﻔﺎﺀ ﻴﻭﻨﺱ ﺍﻟﺼﻔﺎﻭﻱ‬
‫ــــــــــــــــــــــــــــــــــــــــــــــ‬
‫ﺍﻟﻤﻠﺨﺹ‪:‬‬
‫ﻓﻲ ﻫﺫﺍ ﺍﻟﺒﺤﺙ ﺘﻡ ﺘﻭﻀﻴﺢ ﺘﻘﺩﻴﺭ ﻤﻌﻠﻤﺎﺕ ﺍﻻﻨﺤﺩﺍﺭ ﺒﺎﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ‬
‫ﺍﻟﺠﺯﺌﻴﺔ ﻭﻤﻘﺎﺭﻨﺘﻬﺎ ﺒﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻻﻋﺘﻴﺎﺩﻴﺔ ﻤﻥ ﺤﻴﺙ ﻗﺎﺒﻠﻴﺘﻬﺎ ﻋﻠﻰ ﺍﻟﺘﺨﻠﺹ ﻤـﻥ‬
‫ﻤﺸﻜﻠﺔ ﺍﻟﺘﻌﺩﺩ ﺍﻟﺨﻁﻲ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ‪,‬ﺘﺘﻤﻴﺯ ﺍﻟﻁﺭﻴﻘﺔ ﺍﻷﻭﻟﻰ ﺒﺈﻤﻜﺎﻨﻴـﺔ ﺇﻴﺠـﺎﺩ ﺘﺤﻠﻴـل‬
‫ﺍﻻﻨﺤﺩﺍﺭ ﻟﻌﺩﺓ ﻤﺘﻐﻴﺭﺍﺕ ﺍﺴﺘﺠﺎﺒﺔ ﻤﻊ ﻋﺩﺓ ﻤﺘﻐﻴﺭﺍﺕ ﺘﻨﺒﺅﻴـﺔ ﻓـﻲ ﺇﻥ ﻭﺍﺤـﺩ‪،‬ﻭﺘﻡ ﺘﻁﺒﻴـﻕ ﻜﻠﺘـﺎ‬
‫ﺍﻟﻁﺭﻴﻘﺘﻴﻥ ﻋﻠﻰ ﺒﻴﺎﻨﺎﺕ ﺇﻨﺘﺎﺝ ﻤﺎﺩﺓ ﺍﻻﺴﻤﻨﺕ ﺍﺫ ﻭﺼﻔﺕ ﺍﻟﻤﻭﺍﺩ ﺍﻟﺩﺍﺨﻠﺔ ﻓﻲ ﺼﻨﺎﻋﺔ ﻤﺎﺩﺓ ﺍﻻﺴﻤﻨﺕ‬
‫ﺒﺎﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﺇﻤﺎ ﺒﺎﻟﻨﺴﺒﺔ ﺇﻟﻰ ﻤﺘﻐﻴﺭﺍﺕ ﺍﻻﺴﺘﺠﺎﺒﺔ ﻓﺘﻤﺜﻠﺕ ﺒﻤﺎﺩﺓ ﺍﻻﺴﻤﻨﺕ ﻭﺍﻟﻜﻠﻨﻜﺭ ﻭﺍﻟﻐﺒﺎﺭ‬
‫ﻭﺍﻟﻨﻔﺎﻴﺎﺕ ﺍﻟﺼﻠﺒﺔ‪.‬‬

‫‪The Use of Partial Least Squares Method To Remove‬‬

‫‪Multicollinearity‬‬
‫‪Abstract:‬‬

‫‪In this research the estimation of the regression parameters was‬‬

‫‪clarified using least squares and partial least squares compared with the‬‬
‫‪normal method in terms of its ability to be free from the problem of‬‬
‫‪Multicollinearity between the predictive variables, the first method is the‬‬
‫‪possibility of a regression analysis of several response variables with a‬‬
‫‪number of predictive variables at the same time, was the application of‬‬
‫‪both methods on the data production of cement as described in the‬‬
‫‪manufacture Material of cement, either predictive variables for the‬‬
‫‪response variables by cement and clinker, dust and solid waste.‬‬

‫* ‪.‬أﺳﺘﺎذ ﻣﺴﺎﻋﺪ ‪ ،‬آﻠﻴﺔ ﻋﻠﻮم اﻟﺤﺎﺳﺒﺎت واﻟﺮﻳﺎﺿﻴﺎت ‪،‬ﻗﺴﻢ اﻹﺣﺼﺎء واﻟﻤﻌﻠﻮﻣﺎﺗﻴﺔ‬

‫**‪.‬ﻣﺪرس ﻣﺴﺎﻋﺪ ‪ ،‬آﻠﻴﺔ ﻋﻠﻮم اﻟﺤﺎﺳﺒﺎت واﻟﺮﻳﺎﺿﻴﺎت ‪،‬ﻗﺴﻢ اﻹﺣﺼﺎء واﻟﻤﻌﻠﻮﻣﺎﺗﻴﺔ ‪.‬‬
‫*** ﻣﺪرس ﻣﺴﺎﻋﺪ ‪ ،‬آﻠﻴﺔ ﻋﻠﻮم اﻟﺤﺎﺳﺒﺎت واﻟﺮﻳﺎﺿﻴﺎت ‪،‬ﻗﺴﻢ اﻹﺣﺼﺎء واﻟﻤﻌﻠﻮﻣﺎﺗﻴﺔ‬
‫ﺗﺎرﻳﺦ اﻟﺘﺴﻠﻢ ‪ 2009/7/1:‬ــــــــــــــــــــــــــــــــــــــــــــــــــــــــ ﺗﺎرﻳﺦ اﻟﻘﺒﻮل ‪2009/ 12/ 6 :‬‬
 ‫ــــــــــــــــــــــــــــــــــــــــ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ‪...‬‬ ‫]‪[116‬‬

‫ﻤﻘﺩﻤﺔ ‪:‬‬
‫ﻅﻬﺭﺕ ﺍﻟﺤﺎﺠﺔ ﺇﻟﻰ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟـﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴـﺔ))‪Partial Least (PLS‬‬
‫‪ (Squares‬ﻓﻲ ﺍﻟﻌﺩﻴﺩ ﻤﻥ ﺍﻟﺩﺭﺍﺴﺎﺕ ﻤﻨﻬﺎ ﺍﻟﻜﻴﻤﻴﺎﺀ ﻭﺍﻟﻌﻠﻭﻡ ﺍﻻﺠﺘﻤﺎﻋﻴﺔ ﻭﻴﻌﺩ ﺍﻟﺒﺎﺤﺙ ) ; ‪Wold‬‬
‫‪(1966‬ﺍﻭل ﻤﻥ ﻭﺠﺩ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﻭﺃﺼﺒﺤﺕ ﺸﺎﺌﻌﺔ ﺍﻻﺴﺘﺨﺩﺍﻡ ﻓﻲ ﺍﻟﻁﺏ ﻭﺨﺎﺼﺔ ﻓﻲ ﺍﻟﻤﻌﺎﻟﺠﺎﺕ‬
‫ﺍﻟﺴﺭﻴﺭﻴﺔ ﺤﻴﺙ ﻗﻠﺔ ﻋﺩﺩ ﺍﻟﻤﺸﺎﻫﺩﺍﺕ )ﻋﺩﺩ ﺍﻟﻤﺭﻀﻰ( ﻤﻊ ﻜﺜﺭﺓ ﻋﺩﺩ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ) ‪ ( X' S‬ﺍﻟﻤﺘﻤﺜﻠﺔ‬
‫ﺒﺄﻋﺭﺍﺽ ﺍﻟﻤﺭﻴﺽ ﻤﻊ ﻋﺩﺩ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ) ‪ ( Y'S‬ﻭﺍﻟﻤﺘﻤﺜﻠﺔ ﺒﺎﻟﻤﺴﺘﻭﻯ ﺍﻟﺼﺤﻲ ﻟﻠﻤﺭﻴﺽ‬
‫ﻤﻊ ﺘﺤﺴﻴﻥ ﺤﺎﻟﺘﻪ ﺍﻟﺼﺤﻴﺔ ﻭﻨﺘﺎﺌﺞ ﺃﺨﺭﻯ‪.‬ﻜﻤﺎ ﻁﻭﺭ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ )‪. (Friedman ; 1993‬‬
‫ﺇﻥ ﺍﻟﻬﺩﻑ ﻤﻥ ﺃﺴﻠﻭﺏ ﺍﻨﺤﺩﺍﺭ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻫﻭ ﺍﻟﺘﻨﺒـﺅ ﺒﻤﺘﻐﻴـﺭﺍﺕ ﺍﻻﺴـﺘﺠﺎﺒﺔ‬
‫) ‪ ( Y'S‬ﻤﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ) ‪ ( X' S‬ﻭﻤﻥ ﺜﻡ ﻭﺼﻑ ﻫﺫﺍ ﺍﻟﺘﻨﺒﺅ ﻤﻥ ﺨـﻼل ﻋﻼﻗـﺔ ﻤﺘﻤﺜﻠـﺔ‬
‫ﺒﺎﻟﻨﻤﺎﺫﺝ ﺍﻟﺨﻁﻴﺔ ﺍﻟﺘﻲ ﺴﺘﻜﻭﻨﻬﺎ‪.‬‬
‫ﻤﺘﺠﻬﺎ ﻭ ‪ X‬ﻤﺼﻔﻭﻓﺔ ﺫﺍﺕ ﺭﺘﺒﺔ ﻜﺎﻤﻠﺔ ﻓﺎﻥ ﻋﻤﻠﻴﺔ ﺘﻜﻭﻴﻥ ﻨﻤﻭﺫﺝ ﺨﻁﻲ ﻟﻠﺘﻨﺒﺅ‬ ‫ﻋﻨﺩﻤﺎ ﻴﻜﻭﻥ ‪Y‬‬
‫ﻤﻤﻜﻨﺔ ﻭﻴﻤﻜﻥ ﺘﻁﺒﻴﻕ ﺍﻨﺤﺩﺍﺭ ﻤﺘﻌﺩﺩ ﻤﺘﻐﻴﺭﺍﺕ ﺒﺸﻜل ﺍﻋﺘﻴﺎﺩﻱ ‪ ,‬ﻭﻟﻜـﻥ‬ ‫ﺒـ ‪ Y‬ﻤﻥ ﺨﻼل ‪X‬‬
‫ﻋﻨﺩﻤﺎ ﻴﻜﻭﻥ ﻋﺩﺩ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻜﺒﻴﺭﺍ ﻭﻴﻔﻭﻕ ﻋﺩﺩ ﺍﻟﻤﺸﺎﻫﺩﺍﺕ ﻓﺎﻨﻪ ﻓﻲ ﻫﺫﻩ ﺍﻟﺤﺎﻟﺔ ﺴـﺘﻜﻭﻥ‬
‫ﺍﻟﻤﺼﻔﻭﻓﺔ ‪ X‬ﺸﺎﺫﺓ ‪ Singular‬ﻭﻻ ﻴﻤﻜﻥ ﺘﻁﺒﻴﻕ ﺍﻻﻨﺤﺩﺍﺭ ﻟﻤﺘﻌﺩﺩ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻻﻋﺘﻴﺎﺩﻱ ﻭﻫـﺫﺍ‬
‫ﺃﻴﻀﺎ ﺒﺩﻭﺭﻩ ﺴﻴﻅﻬﺭ ﻤﺸﻜﻠﺔ ﺘﻌﺩﺩ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺨﻁﻴﺔ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ‪.‬‬
‫ﻫﻨﺎﻙ ﺤﺎﻟﺘﺎﻥ ﻓﻲ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ‪ ,‬ﺍﻷﻭﻟﻰ ﻭﺍﻟﺘﻲ ﻴﺭﻤﺯ ﻟﻬـﺎ ﺒــ‬
‫ﻤـﻥ‬ ‫)‪ (PLS1‬ﺘﺘﻤﺜل ﺒﺘﻁﺒﻴﻕ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﻟﻠﺘﻨﺒﺅ ﺒﻘﻴﻡ ﻤﺘﻐﻴـﺭ ﻤﻌﺘﻤـﺩ ﻭﺍﺤـﺩ ‪ Y‬ﻤﻘﺎﺒـل ‪K‬‬
‫ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ‪ X‬ﻭﻫﺫﻩ ﺤﺎﻟﺔ ﺨﺎﺼﺔ ‪ ,‬ﻭﺍﻟﺜﺎﻨﻴﺔ ﻫﻲ ﺍﻟﺤﺎﻟﺔ ﺍﻟﻌﺎﻤﺔ ﻴﺭﻤﺯ ﻟﻬﺎ ﺒـ )‪(PLS2‬‬
‫ﺘﺘﻤﺜــل ﺒﺘﻁﺒﻴــﻕ ﻫــﺫﻩ ﺍﻟﻁﺭﻴﻘــﺔ ﻓــﻲ ﺤﺎﻟــﺔ ) ‪ ( j‬ﻤــﻥ ﺍﻟﻤﺘﻐﻴــﺭﺍﺕ ﺍﻟﻤﻌﺘﻤــﺩﺓ‬
‫‪ ( yij‬ﻤــﻊ ) ‪ ( K‬ﻤــﻥ ﺍﻟﻤﺘﻐﻴــﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴــﺔ‬ ‫) ‪,i =1,2,...,n, j = 2,3,...,n‬‬
‫‪ ( xik‬ﻭﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﺘﺘﻁﻠﺏ ﺘﻜﺭﺍﺭﺍ ﻟﻠﻭﺼﻭل ﺇﻟﻰ ﺍﻟﺤل‬ ‫) ‪,i =1,2,...,n, k = 2,3,...,n‬‬
‫ﺍﻟﻨﻬﺎﺌﻲ‪.(Harald & luis ; 1986) .‬‬

‫ﺍﻟﺠﺎﻨﺏ ﺍﻟﻨﻅﺭﻱ‪:‬‬
‫ﺘﻌﺘﻤﺩ ﻁﺭﻴﻘﺔ ﺍﻨﺤﺩﺍﺭ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻋﻠﻰ ﻤﺼﻔﻭﻓﺔ ﺍﻟﺘﺒﺎﻴﻥ ﺍﻟﻤﺸﺘﺭﻙ))‪(cov (X,Y‬‬
‫‪ ,‬ﺇﺫ ﺃﻥ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﺘﺴﻤﺢ ﺒﺘﺤﺩﻴﺩ ﺍﻟﻌﻭﺍﻤل ﻭﺍﻟﺘﻲ ﻫﻲ ﻋﺒﺎﺭﺓ ﻋﻥ ﺘﺭﺍﻜﻴﺏ ﺨﻁﻴـﺔ ﻟﻠﻤﺘﻐﻴـﺭﺍﺕ‬
‫ﺍﻟﺘﻨﺒﺅﻴﺔ ) ‪ , ( X‬ﻭﺘﻌﺭﻑ ﻫﺫﻩ ﺍﻟﻌﻭﺍﻤل ﺒـﺎﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟـﺼﻤﺎﺀ )‪ (Latent Variables‬ﻭﺍﻟﺘـﻲ‬
‫ﺒﺩﻭﺭﻫﺎ ﺘﻜﻭﻥ ﺃﻓﻀل ﻨﻤﻭﺫﺝ ﻟﻠﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ ) ‪,( Y'S‬ﺘﻌـﺩ ﻁﺭﻴﻘـﺔ ﺍﻟﻤﺭﺒﻌـﺎﺕ ﺍﻟـﺼﻐﺭﻯ‬
 ‫]‪[117‬‬ ‫א‬ ‫‪−‬א‬ ‫א‬ ‫א‬ ‫א‬

‫ﺍﻟﺠﺯﺌﻴـﺔ ﻤـﺸﺎﺒﻬﺔ ﺇﻟـﻰ ﻜـل ﻤـﻥ ﺍﻟﻁﺭﺍﺌـﻕ)‪ (Canonical Correlation‬ﻭ) ‪Principal‬‬

‫‪ (Component‬ﻭ)‪ ( Discriminate Analysis‬ﺇﻻ ﺃﻥ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﺘﻔﺭﺽ ﺸـﺭﻭﻁ ﻋﻨـﺩ‬
‫ﺘﻁﺒﻴﻘﻬﺎ ﻤﻨﻬﺎ)‪:(Dante ; 2006‬‬
‫ﻭﺍﻟﻤﺘﻐﻴـﺭﺍﺕ ‪X‬‬ ‫‪Y‬‬ ‫‪-1‬ﺍﻟﻌﻭﺍﻤل ﺃﻭ ﺍﻟﻤﻜﻭﻨﺎﺕ )‪ (Extracted Factors‬ﻟﻜل ﻤﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ‬
‫ﺘﻜﻭﻥ ﻤﺴﺘﺨﻠﺼﺔ ﻤﻥ ﺍﻟﻤﺼﻔﻭﻓﺎﺕ ) ‪ ( Y′Y‬ﻭ) ‪ ( X ′X‬ﻓﻘﻁ ﻭﻟﻴﺱ ﻤﻥ ﺍﻟﻤﺼﻔﻭﻓﺎﺕ ﺍﻟﺘﻲ ﺠﺎﺀﺕ‬
‫ﺃﻱ ﻟﻴﺱ ﻤﻥ ) ‪ ( X′Y‬ﺍﻭ ﻏﻴﺭﻫﺎ‪.‬‬ ‫‪Y‬‬ ‫ﻤﻥ ﺤﺎﺼل ﻀﺭﺏ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ‪ X‬ﻓﻲ‬
‫‪-2‬ﻋﺩﺩ ﺍﻟﻤﻌﺎﺩﻻﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻴﺠﺏ ﺃﻥ ﻻ ﻴﺘﺠﺎﻭﺯ ﺍﻟﺤﺩ ﺍﻷﺩﻨﻰ ﻟﻌﺩﺩ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ ﺍﻭﺍﻟﺘﻨﺒﺅﻴﺔ‪.‬‬
‫ﺒﺎﻟﻨﺴﺒﺔ ﻟﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻓﻬﻲ ﻁﺭﻴﻘﺔ ﻤﻭﺴﻌﺔ ﻟﻼﻨﺤﺩﺍﺭ ﺍﻟﺨﻁﻲ ﺍﻟﻤﺘﻌﺩﺩ ﻭﻟﻜﻥ‬
‫ﺒﺩﻭﻥ ﺍﻟﺸﺭﻁﻴﻥ ﺍﻟﺴﺎﺒﻘﻴﻥ ‪ ,‬ﺍﺫ ﺇﻥ ﺍﻟﻤﻌﺎﺩﻻﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻭﺍﻟﺘﻲ ﻤﺜﻠﺕ ﻤﻥ ﺨﻼل ﺍﻟﻌﻭﺍﻤل ﺍﻟﺘـﻲ ﺘـﻡ‬
‫ﺍﻟﺤﺼﻭل ﻋﻠﻴﻬﺎ ﻤﻥ ﺍﻟﻤﺼﻔﻭﻓﺎﺕ )‪ ( Y′X′XY‬ﻟﻠﺤﺼﻭل ﻋﻠﻰ ﻨﻤـﻭﺫﺝ ﺍﻻﻨﺤـﺩﺍﺭ ﻟﻠﺘــﻨﺒﺅ‬
‫ﺒﻘﻴﻡ ‪. Y‬‬

‫ﺍﻟﺘﺭﻜﻴﺏ ﺃﻭ ﺍﻟﺒﻨﺎﺀ ﺍﻟﺩﺍﺨﻠﻲ ﻟﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ‪:‬‬

‫ﺇﻥ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﺘﺴﻤﻰ ﺒﺎﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺼﻤﺎﺀ ﺍﻭ ﺍﻟﺫﺍﺘﻴﺔ ﺍﺤﻴﺎﻨﹰﺎ ﻭﺫﻟﻙ ﻷﻨﻬـﺎ‬
‫ﺘﻜﻭﻥ ﻤﺸﺎﺒﻬﺔ ﺒﺘﺭﻜﻴﺒﺘﻬﺎ ﻟﻠﻤﺘﺠﻬﺎﺕ ﺍﻟﺫﺍﺘﻴﺔ ﻓﻲ ﺍﻟﺘﺤﻠﻴل ﺍﻟﺫﺍﺘﻲ ﻭﻴﻤﻜﻥ ﻭﺼﻑ ﻫﺫﻩ ﺍﻟﺘﺭﻜﻴﺒﺔ ﺒﺎﻟﺸﻜل‬
‫ﺍﻻﺘﻲ ‪:‬‬
‫ﻟﻴﻜﻥ ﻟﺩﻴﻨﺎ ﻤﺩﻭﺭ ﻤﺼﻔﻭﻓﺔ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ) ' ‪ ( X‬ﻤﻀﺭﻭﺒﺔ ﺒﻤﺘﺠﻪ ﻋﺸﻭﺍﺌﻲ ﻗﺒل ‪ U‬ﺃﻱ ﺃﻥ‪:‬‬
‫‪S‬‬

‫⎤ ‪⎡∑ X i1U i‬‬

‫⎢‬ ‫⎥‬
‫⎥ ‪⎢∑ X i 2U i‬‬
‫⎢ = ‪X ′U‬‬ ‫⎥‬ ‫)‪.... (1‬‬
‫‪⎢ M‬‬ ‫⎥‬
‫⎥ ‪⎢∑ X ipU i‬‬
‫⎣‬ ‫⎦‬
‫ﻭﻤﻥ ﺨﻼﻟﻬﺎ ﻴﻤﻜﻥ ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﻤﺼﻔﻭﻓﺔ ﺍﻟـ ‪ β‬ﻭﺍﻟﺘﻲ ﺘﻭﻀﺢ ﻨﻭﻉ ﻭﺸﻜل ﺍﻟﻌﻼﻗﺔ ﺒـﻴﻥ ﻜـل‬
‫ﻤﺘﻐﻴﺭ ﻤﻌﺘﻤﺩ ﻤﻊ ﻜل ﻤﺘﻐﻴﺭ ﺘﻨﺒﺅﻱ ﺃﻱ ﺃﻥ ‪:‬‬
‫‪Y1 Y2 Y3 L Y j‬‬
‫‪X1‬‬ ‫‪β 11 β 12 β13 L β 1 j‬‬
‫‪X2‬‬ ‫‪β 21 β 22 β 23 L β 2 j‬‬
‫‪X3‬‬ ‫‪β 31 β 32 β 33 L β 3 j‬‬
‫‪M‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪M‬‬
‫‪X K β K 1 β K 2 β K 3 L β Kj‬‬
‫ﺇﻥ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻋﺒﺎﺭﺓ ﻋﻥ ﺘﺭﻜﻴﺒﺔ ﺨﻁﻴﺔ ﻟﻠﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﻟﻤﺼﻔﻭﻓﺔ‬
‫ﺍﻻﺭﺘﺒﺎﻁ ﻭﺍﻟﺘﺒﺎﻴﻥ ﺍﻟﻤﺸﺘﺭﻙ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻭﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ ‪ ,‬ﻭﺍﻟﺠﺯﺀ ﺍﻟﺫﻱ ﺘﻌﺘﻤـﺩ‬

‫‪117‬‬
 ‫ــــــــــــــــــــــــــــــــــــــــ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ‪...‬‬ ‫]‪[118‬‬

‫ﻋﻠﻴﻪ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻓﻲ ﻤﺼﻔﻭﻓﺔ ﺍﻻﺭﺘﺒﺎﻁ ﻭﺍﻟﺘﺒﺎﻴﻥ ﺍﻟﻤﺸﺘﺭﻙ ﻫـﻭ ﺠـﺯﺀ‬
‫‪ Cross Block‬ﺃﻱ ﺍﻻﺭﺘﺒﺎﻁﺎﺕ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻭﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ ‪ ,‬ﻜﻤـﺎ‬ ‫ﺍﻟﺘﻘﺎﻁﻊ‬
‫‪ Factor Scores‬ﻋﻠﻰ ﺸﻜل ﻤﺠﻤﻭﻋﺎﺕ ﺨﻁﻴﺔ ﺒـﻴﻥ ﺍﻟﻤﺘﻐﻴـﺭﺍﺕ‬ ‫ﺘﻘﺩﻡ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﻋﻭﺍﻤل‬
‫ﺍﻟﺘﻨﺒﺅﻴﺔ ﺍﻷﺼﻠﻴﺔ ‪ ,‬ﺒﺫﻟﻙ ﺴﻭﻑ ﻻ ﺘﻜﻭﻥ ﻫﻨﺎﻙ ﺍﺭﺘﺒﺎﻁﺎﺕ ﺒﻴﻥ ﻋﻭﺍﻤل ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﺴﺘﺨﺩﻤﺔ ﻓـﻲ‬
‫ﻨﻤﻭﺫﺝ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺘﻨﺒﺅﻱ ‪.‬‬
‫ﻤﻥ ﺨﻼل ﺍﻟﻌﻼﻗـﺔ ) ; ‪Saikat & Jun‬‬ ‫‪Y‬‬ ‫ﻭﻋﻠﻴﻪ ﻴﻤﻜﻥ ﺇﺠﺭﺍﺀ ﺘﺤﻠﻴل ﻟﻠﻤﺼﻔﻭﻓﺘﻴﻥ ‪, X‬‬
‫‪(2008‬‬

‫‪X=T P′ + K‬‬ ‫)‪…(2‬‬

‫‪Y = U C′ + R‬‬ ‫)‪…(3‬‬
‫ﺤﻴﺙ ﺍﻥ‪:‬‬
‫ٍ‪ = T‬ﻤﺼﻔﻭﻓﺔ ﺍﻟﻌﻭﺍﻤل ﺍﻟﻤﺴﺘﺨﻠﺼﺔ ﻤﻥ ﺍﻟﻤﺼﻔﻭﻓﺔ ‪.X‬‬
‫‪ = P‬ﻫﻭ ﻤﺘﺠﻪ ﻤﺤﻤل ﻟﻠﻤﺼﻔﻭﻓﺔ ‪.X‬‬
‫‪ = U‬ﻤﺼﻔﻭﻓﺔ ﺍﻟﻌﻭﺍﻤل ﺍﻟﻤﺴﺘﺨﻠﺼﺔ ﻤﻥ ﺍﻟﻤﺼﻔﻭﻓﺔ ‪.Y‬‬
‫‪ = C‬ﻤﺘﺠﻪ ﻤﺤﻤل ﻟﻠﻤﺼﻔﻭﻓﺔ ‪.Y‬‬
‫ﻭﺍﻥ ﻋﻤﻠﻴﺔ ﺍﺴﺘﺨﻼﺹ ﻋﻭﺍﻤل ﺍﻟﻤﺼﻔﻭﻓﺔ ‪ X‬ﺘﺘﻤﺜل ﺒﺈﻴﺠﺎﺩ ﺘﺭﻜﻴﺒﺔ ﺨﻁﻴﺔ ﻤﻥ ﺃﻋﻤﺩﺓ ﺍﻟﻤﺼﻔﻭﻓﺔ‬
‫‪ , X‬ﺃﻱ ﺇﻴﺠﺎﺩ ﻤﺘﺠﻪ ﺍﺒﺘﺩﺍﺌﻲ ﻴﺘﻡ ﻀﺭﺒﻪ ﺒﺎﻟﻤﺼﻔﻭﻓﺔ ‪ X‬ﻹﻴﺠﺎﺩ ﺍﻟﺘﺭﺍﻜﻴﺏ ﺍﻟﺨﻁﻴﺔ ‪ t‬ﺃﻱ ﺍﻥ ‪:‬‬
‫‪t = XW‬‬
‫ﺤﻴﺙ ﺍﻥ ‪ W‬ﻫﻭ ﻤﺘﺠﻪ ﻟﻘﻴﻡ ﻋﺸﻭﺍﺌﻴﺔ ﺃﻭ ﻫﻭ ﺃﻭل ﻤﺘﺠﻪ ﻤﻤﻴـﺯ ﺍﻟﻤﻘﺎﺒـل ﻷﻭل ﻗﻴﻤـﺔ ﻤﻤﻴـﺯﺓ‬
‫ﻭﺍﻟﻤﺘﻤﺜﻠـﺔ‬ ‫‪Y‬‬ ‫ﻟﻠﻤﺼﻔﻭﻓﺔ )‪ , ( X′YY′X‬ﺒﺎﻟﻁﺭﻴﻘﺔ ﻨﻔﺴﻬﺎ ﻓﺈﻥ ﺍﻟﺘﺭﻜﻴﺒﺔ ﺍﻟﺨﻁﻴـﺔ ﻟﻠﻤـﺼﻔﻭﻓﺔ‬
‫ﺒﺎﻟﻤﺘﺠﻪ ‪ U‬ﻜﻤﺎ ﻴﺎﺘﻲ ‪:‬‬
‫‪U= YC‬‬
‫ﺤﻴﺙ ﺃﻥ ‪ C‬ﻫﻭ ﺃﻭل ﻤﺘﺠﻪ ﻤﻤﻴﺯ ﻤﻘﺎﺒل ﺇﻟﻰ ﺃﻭل ﺠﺫﺭ ﻤﻤﻴـﺯ ﻟﻠﻤـﺼﻔﻭﻓﺔ ) ‪ ( Y′XX′Y‬ﻤـﻊ‬
‫ﺍﻟﻌـﻠﻡ ﺃﻥ ) ‪ ( X ′Y‬ﺘﻤﺜل ﻤﺼﻔﻭﻓﺔ ﺍﻟﺘﺒﺎﻴﻥ ﺍﻟﻤﺸﺘﺭﻙ ﺒﻴﻥ ‪ X‬ﻭ ‪. Y‬‬
‫ﺇﻥ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﺘﺒﺤﺙ ﻋﻥ ﻤﺠﻤﻭﻋﺔ ﻤﻥ ﺍﻟﻤﻜﻭﻨﺎﺕ ﻭﺍﻟﺘﻲ ﺘﺩﻋﻰ ﺒﺎﻟﻤﺘﺠﻬﺎﺕ ﺍﻟﺼﻤﺎﺀ ) ‪Latent‬‬
‫‪.‬‬ ‫‪Y‬‬ ‫‪ (Vectors‬ﻭﺍﻥ ﻫﺫﻩ ﺍﻟﻤﺘﺠﻬﺎﺕ ﺘﻭﻀﺢ ﻗﺩﺭ ﺍﻹﻤﻜﺎﻥ ﺍﻟﺘﺒﺎﻴﻥ ﺍﻟﻤﺸﺘﺭﻙ ﺒﻴﻥ ‪ X‬ﻭ‬
‫ﻭﻴﻤﻜﻥ ﺘﺤﻠﻴل ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﺴﺘﻘﻠﺔ ﻤﻥ ﺨﻼل ﺍﻟﻤﻌﺎﺩﻟﺔ ﺍﻻﺘﻴﺔ ‪:‬‬
‫‪X = T P′‬‬
‫ﻫﻲ ﻋﺒﺎﺭﺓ ﻋﻥ ﻤﺠﻤﻭﻋﺔ ﺨﻁﻴﺔ ﻤﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻭﻟﻜﻥ ﻋﻠﻰ ﺸـﻜل ﻋﻭﺍﻤـل‬ ‫ﺤﻴﺙ ‪T‬‬
‫ﻤﺘﻌﺎﻤﺩﺓ ‪ ,‬ﺃﻱ ﺃﻥ ﻜل ﻋﻤﻭﺩ ﻴﺤﻭﻱ ﺠﻤﻴﻊ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﺴﺘﻘﻠﺔ ﺍﻟﻤﻭﺠﻭﺩﺓ ﻓﻲ ‪ , X‬ﻭﻟﻜﻥ ﻋﻠـﻰ‬
‫]‪[119‬‬ ‫א‬ ‫‪−‬א‬ ‫א‬ ‫א‬ ‫א‬

‫ﻓﻬﻭ ﻤﺘﺠﻪ ﻤﺤﻤل ﻭﺍﻟﻤﻘﺼﻭﺩ ﺒـﻪ ﻋﺒـﺎﺭﺓ ﻋـﻥ‬ ‫‪P‬‬ ‫ﺸﻜل ﻤﺠﻤﻭﻋﺔ ﺨﻁﻴﺔ ﻤﻥ ﺍﻷﻭﺯﺍﻥ ‪ ,‬ﺃﻤﺎ‬
‫ﻭﺍﻟﻤﺼﻔﻭﻓﺔ ﺍﻷﺼﻠﻴﺔ ﻟﻠﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﺴﺘﻘﻠﺔ ﺃﻱ ﺍﻥ ‪:‬‬ ‫‪t‬‬ ‫ﻤﺠﻤﻭﻋﺔ ﺨﻁﻴﺔ ﺒﻴﻥ ﺍﻟﻌﻭﺍﻤل ﺍﻟﻤﺘﻌﺎﻤﺩﺓ‬
‫‪P = X′ t‬‬
‫ﺤﻴﺙ ﺍﻥ ‪ t‬ﻫﻭ ﻋﻤﻭﺩ ﻤﻥ ﺃﻋﻤﺩﺓ ﺍﻟﻤﺼﻔﻭﻓﺔ ‪T‬‬
‫‪T′T = Ι‬‬
‫‪P′P ≠ Ι‬‬
‫ﻭﻴﺘﻡ ﺘﻜـﺭﺍﺭ ﺃﻭ ﺇﻋـﺎﺩﺓ ﻫـﺫﺍ‬ ‫‪Y‬‬ ‫ﻭﺒﻌﺩ ﺇﻴﺠﺎﺩ ﺃﻭل ﻤﺘﺠﻪ ﺫﺍﺘﻲ ‪ ,‬ﻴﺘﻡ ﻁﺭﺤﻪ ﻤﻥ ﻜل ﻤﻥ ‪ X‬ﻭ‬
‫ﺍﻹﺠﺭﺍﺀ ﺤﺘﻰ ﺘﺼﺒﺢ ‪ X‬ﻤﺼﻔﻭﻓﺔ ﺼﻔﺭﻴﺔ ‪.‬‬

‫ﺨﻭﺍﺭﺯﻤﻴﺔ ﻁﺭﻴﻘﺔ ﺍﻨﺤﺩﺍﺭ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ‪:‬‬

‫ﻴﻤﻜﻥ ﺘﻭﻀﻴﺢ ﺁﻟﻴﺔ ﻋﻤل ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻤﻥ ﺤﻴﺙ ﺤﺴﺎﺏ ﻤﻌﻠﻤﺎﺕ ﺍﻻﻨﺤﺩﺍﺭ‬
‫‪ ,‬ﻭﻜﺨﻁﻭﺓ ﺃﻭﻟﻴـﺔ ﻴـﺘﻡ ﺘﺤﻭﻴـل‬ ‫‪P‬‬ ‫ﻭﺍﻷﻋﻤﺩﺓ ﺍﻟﻤﺤﻤﻠﺔ‬ ‫ﻭﻤﺼﻔﻭﻓﺔ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺼﻤﺎﺀ ‪T‬‬
‫ﺇﻟﻰ ﺍﻟﺼﻴﻐﺔ ﺍﻟﻘﻴــﺎﺴﻴﺔ‬ ‫‪Y‬‬ ‫ﻤﺼﻔﻭﻓﺔ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ‪ X‬ﻭﻤﺼﻔﻭﻓﺔ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ‬
‫ﻋﻠﻰ ﺍﻟﺘﻭﺍﻟﻲ ‪ ,‬ﻭﺨﻁﻭﺍﺕ ﻫﺫﻩ ﺍﻟﺨﻭﺍﺭﺯﻤﻴﺔ ﻜﻤﺎ ﻴﺎﺘﻲ ‪:‬‬ ‫ﺒﺤﻴﺙ ﺘﺼﺒﺢ ‪ E‬ﻭ ‪F‬‬
‫‪ -1‬ﺍﺨﺫ ﻗﻴﻡ ﺍﺒﺘﺩﺍﺌﻴﺔ ﻋﺸﻭﺍﺌﻴﺔ ﻟﻠﻤﺘﺠﻪ ‪.U‬‬
‫‪ -2‬ﺇﻴﺠﺎﺩ ﺍﻟﻤﺘﺠﻪ ‪ W‬ﻤﻥ ﺨﻼل ﺍﻟﻌﻼﻗﺔ‪:‬‬
‫‪W=E′U‬‬
‫ﻟﻠﻤﺘﻐﻴـﺭﺍﺕ ﺍﻟﻤﺴﺘﺨﻠـﺼﺔ ﻤـﻥ‬ ‫‪ -3‬ﺇﻴﺠﺎﺩ ﺍﻟﻤﺘﺠﻪ ‪ t‬ﻭﻫﻭ ﺍﺤـﺩ ﺃﻋﻤـﺩﺓ ﺍﻟﻤـﺼﻔﻭﻓﺔ ‪T‬‬
‫ﺍﻟﻤﺼﻔﻭﻓﺔ ‪ X‬ﺃﻱ ﺃﻥ‪:‬‬
‫‪t old = EW‬‬
‫ﺃﻱ ﺃﻥ ‪:‬‬ ‫‪t‬‬ ‫ﻋﻠﻰ ﻁﻭل ﺍﻟﻤﺘﺠﻪ‬ ‫‪t‬‬ ‫‪-4‬ﻗﺴﻤﺔ ﺍﻟﻤﺘﺠﻪ‬
‫‪t old‬‬
‫= ‪t new‬‬
‫‪t‬‬
‫ﺤﻴﺙ ﺃﻥ‬
‫‪2‬‬ ‫‪2‬‬ ‫‪2‬‬ ‫‪2‬‬
‫‪t‬‬ ‫=‬ ‫‪t1 + t 2 + t 3 + ....... + t n‬‬

‫‪-5‬ﺇﻴﺠﺎﺩ ﺍﻟﻤﺘﺠﻪ ﺍﻟﻤﻭﺯﻭﻥ ‪ C‬ﻟﻠﻤﺼﻔﻭﻓﺔ ‪: Y‬‬

‫‪C= F′ t‬‬
‫‪-6‬ﺇﻴﺠﺎﺩ ﺍﻟﻤﺘﺠﻪ ‪ U‬ﻭﻫﻭ ﺍﺤﺩ ﺃﻋﻤﺩﺓ ﺍﻟﻤﺼﻔﻭﻓﺔ ‪ U‬ﻟﻠﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﺴﺘﺨﻠﺼﺔ ﻟﻠﻤﺼﻔﻭﻓﺔ ‪Y‬‬

‫‪119‬‬
 ‫ــــــــــــــــــــــــــــــــــــــــ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ‪...‬‬ ‫]‪[120‬‬

‫‪Uold = FC‬‬
‫‪-7‬ﻗﺴﻤﺔ ﺍﻟﻤﺘﺠﻪ ‪ U‬ﻋﻠﻰ ﻁﻭل ﺍﻟﻤﺘﺠﻪ ‪ U‬ﻭﻜﻤﺎ ﻴﺎﺘﻲ ‪:‬‬
‫‪U old‬‬
‫= ‪U new‬‬
‫‪U‬‬
‫ﺤﻴﺙ ﺃﻥ‬
‫‪2‬‬ ‫‪2‬‬ ‫‪2‬‬ ‫‪2‬‬
‫‪U‬‬ ‫=‬ ‫‪U1 +U 2 +U 3‬‬ ‫‪+ ....... + U n‬‬

‫‪-8‬ﻤﻘﺎﺭﻨﺔ ﻗﻴﻡ ﺍﻟﻤﺘﺠﻪ ‪ U‬ﻓﻲ ﺍﻟﺨﻁﻭﺓ )‪ (7‬ﻤﻊ ﻗﻴﻡ ﺍﻟﻤﺘﺠﻪ ‪ t‬ﻓﻲ ﺍﻟﺨﻁﻭﺓ )‪ (4‬ﻓﺈﺫﺍ ﻜﺎﻨﺕ ﺍﻟﻘـﻴﻡ‬
‫ﻤﺘﻘﺎﺭﺒﺔ ﻴﺘﻡ ﺤﺴﺎﺏ ﻗﻴﻡ ‪ β‬ﻭﻓﻲ ﺤﺎﻟﺔ ﻋـﺩﻡ ﺍﻟﺘﻘـﺎﺭﺏ ﻴـﺘﻡ ﺤـﺴـﺎﺏ ﻤﺘﺠـﻪ ﺍﻟﻤﺤﻤـل ‪P‬‬
‫ﻟﻠﻤﺼﻔﻭﻓﺔ ‪ X‬ﻤﻥ ﺨﻼل ﺍﻟﻤﻌﺎﺩﻟﺔ ﺍﻻﺘﻴﺔ‪:‬‬
‫‪P = E′t‬‬
‫ﺒﻌﺩ ﺫﻟﻙ ﻴﺘﻡ ﻁﺭﺡ ﺘﺄﺜﻴﺭ ﺍﻟﻤﺘﺠﻪ ‪ (Partial Out) t‬ﻤﻥ ﻜل ﻤﻥ ﺍﻟﻤﺼﻔﻭﻓﺘﻴﻥ ‪ E‬ﻭ ‪ F‬ﻭﻜﻤﺎ ﻴﻠﻲ‪:‬‬
‫‪E 1 = E − tP ′‬‬
‫‪F1 = F − b t C ′‬‬

‫ﻭﻤﻥ ﺍﻟﻤﻌﺎﺩﻟﺘﻴﻥ ﻤﻥ ﺍﻟﺨﻁﻭﺘﻴﻥ )‪ (5‬ﻭ )‪ (8‬ﻨﺠﺩ ‪:‬‬

‫‪E 1 = E − t( E ′t ) ′‬‬
‫‪= E − t t ′E‬‬

‫‪F1 = F − b t( F ′ t ) ′‬‬
‫‪= F − b t t ′F‬‬

‫ﺒﻌﺩ ﺫﻟﻙ ﺘﻌﺎﺩ ﺍﻟﺨﻭﺍﺭﺯﻤﻴﺔ ﻤـﻥ ﺠﺩﻴـﺩ ﺒﻨﻔﺱ ﺍﻟﺨﻁـﻭﺍﺕ ﻭﻟﻜــﻥ ﺒﺎﺴـﺘﺨﺩﺍﻡ ﺍﻟﻤـﺼﻔﻭﻓﺘﻴﻥ‬
‫ﻤﻊ ﺍﻟﻌﻠﻡ ﺃﻥ ﻓﻲ ﻜل ﻋﻤﻠﻴﺔ ﺘﻜﺭﺍﺭ ﺘﺘﻡ ﻤﻘﺎﺭﻨﺔ ﻗـﻴﻡ ﺍﻟﻤﺘﺠـﻪ )‪ (t‬ﻟﻠـﺩﻭﺭﺓ‬ ‫ﺍﻟﺠﺩﻴﺩﺘﻴﻥ ‪ E1‬ﻭ ‪F 1‬‬
‫ﺍﻷﺨﻴﺭﺓ ﻤﻊ ﻗﻴﻡ ﺍﻟﻤﺘﺠﻪ )‪ (t‬ﻟﻠﺩﻭﺭﺓ ﺍﻟﺴﺎﺒﻘﺔ ‪ ,‬ﻭﺘﺴﺘﻤﺭ ﻋﻤﻠﻴﺔ ﺍﻟﺘﻜﺭﺍﺭ ﺇﻟﻰ ﺃﻥ ﻴﺘﻡ ﺍﻟﺤﺼﻭل ﻋﻠـﻰ‬
‫)ﻭﺍﻟﺘﻲ ﻋﺩﺩﻫﺎ ﻤﺴﺎﻭ ﺇﻟﻰ ﺭﺘﺒﺔ ﺍﻟﻤﺼﻔﻭﻓﺔ ‪ ( X‬ﻭﻤﻥ ﺜﻡ ﻴﺘﻡ ﺇﻴﺠـﺎﺩ‬ ‫ﺠﻤﻴﻊ ﺃﻋﻤﺩﺓ ﺍﻟﻤﺼﻔﻭﻓﺔ ‪T‬‬
‫ﻤﺼﻔﻭﻓﺔ ﺍﻟﻤﻌﻠﻤﺎﺕ ‪ β‬ﻭﻜﻤﺎ ﻴﺎﺘﻲ ‪:‬‬
‫‪β = E ′U (T ′EE ′U ) −1 T ′F‬‬
 ‫]‪[121‬‬ ‫א‬ ‫‪−‬א‬ ‫א‬ ‫א‬ ‫א‬

‫ﺍﻟﺠﺎﻨﺏ ﺍﻟﻌﻤﻠﻲ ‪:‬‬

‫ﺘﻡ ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﺍﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﻤﺴﺘﺨﺩﻤﺔ ﻓﻲ ﻫﺫﺍ ﺍﻟﺒﺤﺙ ﻤﻥ ﻤﻌﻤـل ﺍﺴـﻤﻨﺕ ﻜﺭﻜـﻭﻙ‬
‫ﻭﻟﻠﻔﺘﺭﺓ )‪ (2006 - 2002‬ﺍﺫ ﺘﻤﺜﻠﺕ ﺍﻟﻌﻤﻠﻴﺔ ﺍﻹﻨﺘﺎﺠﻴﺔ ﻤﻥ ﺃﺭﺒﻌﺔ ﻋﻨﺎﺼﺭ ﺃﺴﺎﺴﻴﺔ ﺘﻤﺜـل ﻤـﻭﺍﺩ‬
‫ﺃﻭﻟﻴﺔ )ﻤﺩﺨﻼﺕ( ﻭﺍﻟﺘﻲ ﺍﻋﺘﺒﺭﺕ ﻤﺘﻐﻴﺭﺍﺕ ﺘﻨﺒﺅﻴﺔ ﻹﻨﺘﺎﺝ ﻤﺎﺩﺓ ﺍﻻﺴـﻤﻨﺕ ﻓـﻀﻼ ﻋـﻥ ﺃﺭﺒﻌـﺔ‬
‫ﻤﺨﺭﺠﺎﺕ ﻤﻊ ﺍﻻﺴﻤﻨﺕ ﻭﺍﻟﺘﻲ ﺘﻌﺘﺒﺭ ﻤﺘﻐﻴﺭﺍﺕ ﻤﻌﺘﻤﺩﺓ ‪ ,‬ﺒﺫﻟﻙ ﺴﻴﻜﻭﻥ ﻫﻨﺎﻙ ﺃﺭﺒﻌـﺔ ﻤﺘﻐﻴـﺭﺍﺕ‬
‫ﺘﻨﺒﺅﻴﺔ )ﻤﺩﺨﻼﺕ( ﻭﺃﺭﺒﻌﺔ ﻤﺘﻐﻴﺭﺍﺕ ﻤﻌﺘﻤﺩﺓ )ﻤﺨﺭﺠﺎﺕ( ‪ ,‬ﻭﻜﻤﺎ ﻤﻭﻀﺤﺔ ﺒﺎﻟﺠﺩﻭل ﺍﻷﺘﻲ ‪:‬‬

‫ﺍﻟﺠﺩﻭل )‪ (1‬ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻭﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ ﻟﻌﻤﻠﻴﺔ ﺇﻨﺘﺎﺝ ﺍﻻﺴﻤﻨﺕ‬

‫ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ‬ ‫ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ‬

‫‪ : X1‬ﺍﻟﻁﺎﻗﺔ ﺍﻟﻜﻬﺭﺒﺎﺌﻴﺔ ‪/‬‬ ‫‪ : Y1‬ﻜﻤﻴﺔ ﺇﻨﺘﺎﺝ ﺍﻻﺴﻤﻨﺕ ‪ /‬ﻁﻥ‬
‫‪ : X2‬ﺍﻟﻨﻔﻁ ‪ /‬ﻁﻥ‬ ‫‪ : Y2‬ﺇﻨﺘﺎﺝ ﺍﻟﻜﻠﻨﻜﺭ ‪ /‬ﻁﻥ‬
‫‪ : X3‬ﺍﻟﺤﺠﺭ ‪ /‬ﻁﻥ‬ ‫‪ : Y3‬ﺍﻟﻨﻔﺎﻴﺎﺕ ﺍﻟﺼﻠﺒﺔ ‪ /‬ﻁﻥ‬
‫‪ : X 4‬ﺍﻟﻁﻴﻥ ‪ /‬ﻁﻥ‬ ‫‪3‬‬
‫‪ : Y4‬ﻜﻤﻴﺎﺕ ﺍﻟﻐﺒﺎﺭ ﺍﻟﻤﻨﺒﻌﺜﺔ ‪ /‬ﻡ‬

‫‪ -1‬ﺃﺴﻠﻭﺏ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻻﻋﺘﻴﺎﺩﻴﺔ )‪:(OLS‬‬

‫ﺘﻡ ﺘﺤﻠﻴل ﺍﻟﺒﻴﺎﻨﺎﺕ ﺃﻋﻼﻩ ﺒﺎﺴﺘﺨﺩﺍﻡ ﺃﺴﻠﻭﺏ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻻﻋﺘﻴﺎﺩﻴﺔ ﺒﺎﻨﺤﺩﺍﺭ ﻜـل‬
‫ﻤﺘﻐﻴﺭ ﻤﻌﺘﻤﺩ ﻋﻠﻰ ﺤﺩﺓ ﻤﻊ ﺒﻘﻴﺔ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﺃﻱ ﺍﻨﻪ ﺘﻡ ﺘﺤﻠﻴل ﺍﻨﺤﺩﺍﺭ )‪(y1/x1,x2,x3,x4‬‬
‫ﻭ)‪ (y2/x1,x2,x3,x4‬ﻭ)‪ (y3/x1,x2,x3,x4‬ﻭ)‪ (y4/x1,x2,x3,x4‬ﻭﻜﺎﻨﺕ ﻨﺘﺎﺌﺞ ﺍﻟﺘﺤﻠﻴـل ﺒﺎﻟـﺸﻜل‬
‫ﺍﻻﺘﻲ ‪:‬‬

‫‪-1‬ﺘﺤﻠﻴل ﺍﻨﺤﺩﺍﺭ )‪: (y1/x1, x2,x3,x4‬‬

‫‪Y1 = – 0.000736 + 1.82 X1 – 0.72 X2 – 0.108 X3‬‬ ‫)‪…..(4‬‬

‫ﺍﻟﺠﺩﻭل )‪ (2‬ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻹﻨﺘﺎﺝ ﺍﻻﺴﻤﻨﺕ‬

‫‪Source‬‬ ‫‪D.F‬‬ ‫‪S.S‬‬ ‫‪M.S‬‬ ‫‪F‬‬ ‫‪P‬‬

‫‪Regression‬‬ ‫‪3‬‬ ‫‪5.5079‬‬ ‫‪1.836‬‬
‫‪9180‬‬ ‫‪0.007‬‬
‫‪Residual‬‬ ‫‪1‬‬ ‫‪0.0002‬‬ ‫‪0.0002‬‬
‫‪Total‬‬ ‫‪4‬‬ ‫‪5.508‬‬
‫‪R2 = 100 %‬‬ ‫‪2‬‬
‫‪R (adj) = 100 %‬‬

‫‪121‬‬
 ‫ــــــــــــــــــــــــــــــــــــــــ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ‪...‬‬ ‫]‪[122‬‬

‫ﺍﻟﺠﺩﻭل )‪ (3‬ﻗﻴﻡ ﻤﻌﻠﻤﺎﺕ ﺍﻻﻨﺤﺩﺍﺭ ‪ β‬ﻭﺍﻟـ ‪ VIF‬ﻭﺍﻟﻤﺨﺘﺒﺭ ﺍﻹﺤﺼﺎﺌﻲ ‪T‬‬

‫‪Predictor‬‬ ‫‪D.F‬‬ ‫‪Coef.‬‬ ‫‪SE Coef.‬‬ ‫‪T‬‬ ‫‪P‬‬ ‫‪VIF‬‬

‫‪Constant‬‬ ‫‪--‬‬ ‫‪–0.000736‬‬ ‫‪0.00796‬‬ ‫‪– 0.09‬‬ ‫‪0.94‬‬ ‫‪---‬‬
‫‪X1‬‬ ‫‪1‬‬ ‫‪1.824‬‬ ‫‪2.022‬‬ ‫‪0.90‬‬ ‫‪0.53‬‬ ‫‪145523.8‬‬
‫‪X2‬‬ ‫‪1‬‬ ‫‪– 0.72‬‬ ‫‪1.858‬‬ ‫‪– 0.39‬‬ ‫‪0.765‬‬ ‫‪123504.4‬‬
‫‪X3‬‬ ‫‪1‬‬ ‫‪– 1083‬‬ ‫‪0.171‬‬ ‫‪– 0.63‬‬ ‫‪0.641‬‬ ‫‪963.4‬‬

‫‪-2‬ﺘﺤﻠﻴل ﺍﻨﺤﺩﺍﺭ )‪: (y2/x1,x2,x3,x4‬‬

‫‪Y2 = 0.0000 + 0.0000 X1 – 0.0000 X2 +1.00 X3‬‬ ‫)‪…..(5‬‬

‫ﺍﻟﺠﺩﻭل )‪ (4‬ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻹﻨﺘﺎﺝ ﺍﻟﻜﻠﻨﻜﺭ‬

‫‪Source‬‬ ‫‪D.F‬‬ ‫‪S.S‬‬ ‫‪M.S‬‬ ‫‪F‬‬ ‫‪P‬‬

‫‪Regression‬‬ ‫‪3‬‬ ‫‪5.0993‬‬ ‫‪1.6998‬‬
‫*‬ ‫*‬
‫‪Residual‬‬ ‫‪1‬‬ ‫‪0.0000‬‬ ‫‪0.0000‬‬
‫‪Total‬‬ ‫‪4‬‬ ‫‪5.0993‬‬
‫‪R2 = 100 %‬‬ ‫‪2‬‬
‫‪R (adj) = 100 %‬‬

‫ﺍﻟﺠﺩﻭل )‪ (5‬ﻗﻴﻤﺎ ﻤﻌﻠﻤﺎﺕ ﺍﻻﻨﺤﺩﺍﺭ ‪ β‬ﻭﺍﻟـ ‪ VIF‬ﻭﺍﻟﻤﺨﺘﺒﺭ ﺍﻹﺤﺼﺎﺌﻲ ‪T‬‬

‫‪Predictor‬‬ ‫‪D.F‬‬ ‫‪Coef.‬‬ ‫‪SE Coef.‬‬ ‫‪T‬‬ ‫‪P‬‬ ‫‪VIF‬‬

‫‪Constant‬‬ ‫‪--‬‬ ‫‪0.0000‬‬ ‫‪0.0000‬‬ ‫*‬ ‫*‬ ‫‪---‬‬
‫‪X1‬‬ ‫‪1‬‬ ‫‪0.0000‬‬ ‫‪0.0000‬‬ ‫*‬ ‫*‬ ‫‪145523.8‬‬
‫‪X2‬‬ ‫‪1‬‬ ‫‪– 0.0000‬‬ ‫‪0.0000‬‬ ‫*‬ ‫*‬ ‫‪123504.4‬‬
‫‪X3‬‬ ‫‪1‬‬ ‫‪1.000‬‬ ‫‪0.0000‬‬ ‫*‬ ‫*‬ ‫‪963.4‬‬

‫‪-3‬ﺘﺤﻠﻴل ﺍﻨﺤﺩﺍﺭ)‪:(y3/x1,x2,x3,x4‬‬
‫‪Y3 = - 0.00000414 + 0.00385 X1 – 0.00356X2 +1.00 X3‬‬ ‫)‪…..(6‬‬
‫ﺍﻟﺠﺩﻭل )‪ (6‬ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻟﻜﻤﻴﺎﺕ ﺍﻟﻨﻔﺎﻴﺎﺕ ﺍﻟﺼﻠﺒﺔ‬

‫‪Source‬‬ ‫‪D.F‬‬ ‫‪S.S‬‬ ‫‪M.S‬‬ ‫‪F‬‬ ‫‪P‬‬

‫‪Regression‬‬ ‫‪3‬‬ ‫‪5.0994‬‬ ‫‪1.6998‬‬
‫‪169980‬‬ ‫‪0.000‬‬
‫‪Residual‬‬ ‫‪1‬‬ ‫‪0.0000‬‬ ‫‪0.00001‬‬
‫‪Total‬‬ ‫‪4‬‬ ‫‪5.0994‬‬
‫‪R2 = 100 %‬‬ ‫‪2‬‬
‫‪R (adj) = 100 %‬‬
 ‫]‪[123‬‬ ‫א‬ ‫‪−‬א‬ ‫א‬ ‫א‬ ‫א‬

‫ﺍﻟﺠﺩﻭل )‪ (7‬ﻗﻴﻡ ﻤﻌﻠﻤﺎﺕ ﺍﻻﻨﺤﺩﺍﺭ ‪ β‬ﻭﺍﻟـ ‪ VIF‬ﻭﺍﻟﻤﺨﺘﺒﺭ ﺍﻹﺤﺼﺎﺌﻲ ‪T‬‬

‫‪Predictor D.F‬‬ ‫‪Coef.‬‬ ‫‪SE Coef.‬‬ ‫‪T‬‬ ‫‪P‬‬ ‫‪VIF‬‬

‫‪Constant -- -0.00000414‬‬ ‫‪0.00002868 -0.14‬‬ ‫‪0.909‬‬ ‫‪---‬‬
‫‪X1‬‬ ‫‪1‬‬ ‫‪0.003846‬‬ ‫‪0.007281‬‬ ‫‪0.53‬‬ ‫‪0.691 145523.8‬‬
‫‪X2‬‬ ‫‪1‬‬ ‫‪– 0.00356‬‬ ‫‪0.006691‬‬ ‫‪-0.14‬‬ ‫‪0.689 123504.4‬‬
‫‪X3‬‬ ‫‪1‬‬ ‫‪0.999708‬‬ ‫‪0.000616 1623.58‬‬ ‫‪0.000‬‬ ‫‪963.4‬‬

‫‪-4‬ﺘﺤﻠﻴل ﺍﻨﺤﺩﺍﺭ )‪:(y4/x1,x2,x3,x4‬‬

‫‪Y4 = - 0.431 + 152 X1 + 141X2 +11.00 X3‬‬ ‫)‪…..(7‬‬

‫ﺍﻟﺠﺩﻭل )‪ (8‬ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻟﻜﻤﻴﺎﺕ ﺍﻟﻐﺒﺎﺭ ﺍﻟﻤﻨﺒﻌﺜﺔ‬

‫‪Source‬‬ ‫‪D.F‬‬ ‫‪S.S‬‬ ‫‪M.S‬‬ ‫‪F‬‬ ‫‪P‬‬

‫‪Regression‬‬ ‫‪3‬‬ ‫‪1.424‬‬ ‫‪0.4747‬‬
‫‪1.66‬‬ ‫‪0.505‬‬
‫‪Residual‬‬ ‫‪1‬‬ ‫‪0.2833‬‬ ‫‪0.2853‬‬
‫‪Total‬‬ ‫‪4‬‬ ‫‪1.7093‬‬
‫‪R2 = 100 %‬‬ ‫‪2‬‬
‫‪R (adj) = 100 %‬‬

‫ﺍﻟﺠﺩﻭل )‪ (9‬ﻗﻴﻡ ﻤﻌﻠﻤﺎﺕ ﺍﻻﻨﺤﺩﺍﺭ ‪ β‬ﻭﺍﻟـ ‪ VIF‬ﻭﺍﻟﻤﺨﺘﺒﺭ ﺍﻹﺤﺼﺎﺌﻲ ‪T‬‬

‫‪Predictor D.F‬‬ ‫‪Coef.‬‬ ‫‪SE Coef.‬‬ ‫‪T‬‬ ‫‪P‬‬ ‫‪VIF‬‬

‫‪Constant --‬‬ ‫‪-0.4306‬‬ ‫‪0.3420‬‬ ‫‪-1.26‬‬ ‫‪0.427‬‬ ‫‪---‬‬
‫‪X1‬‬ ‫‪1‬‬ ‫‪-151.78‬‬ ‫‪86.83‬‬ ‫‪-1.75‬‬ ‫‪0.331 145523.8‬‬
‫‪X2‬‬ ‫‪1‬‬ ‫‪141.13‬‬ ‫‪79.80‬‬ ‫‪1.77‬‬ ‫‪0.328 123504.4‬‬
‫‪X3‬‬ ‫‪1‬‬ ‫‪11.011‬‬ ‫‪7.343‬‬ ‫‪1.50‬‬ ‫‪0.374‬‬ ‫‪963.4‬‬

‫ﻤﻥ ﺨﻼل ﻤﻼﺤﻅﺔ ﺘﺤﻠﻴل ﺍﻻﻨﺤﺩﺍﺭ ﻓﻲ ﺍﻟﺠﺩﺍﻭل )‪ (9) , (7) , (5) , (3‬ﺘﻡ ﺤـﺫﻑ ﺍﻟﻤﺘﻐﻴـﺭ‬
‫ﺍﻟﺭﺍﺒﻊ )‪ (X4‬ﻭﺍﻟﺴﺒﺏ ﻓﻲ ﺫﻟﻙ ﻴﻌﻭﺩ ﺇﻟﻰ ﻗﻭﺓ ﺍﻟﺘﺩﺍﺨل ﺍﻟﺨﻁﻲ ﻟﻬﺫﺍ ﺍﻟﻤﺘﻐﻴﺭ ﻤﻊ ﺒﻘﻴﺔ ﺍﻟﻤﺘﻐﻴـﺭﺍﺕ‬
‫ﺍﻟﺘﻨﺒﺅﻴﺔ ﻤﻤﺎ ﻴﻌﻁﻲ ﻀﻌﻔﺎ ﻟﻠﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﺩﺭﻭﺴﺔ ﻋﻠﻤﺎ ﺃﻥ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺭﺍﺒﻊ ﺍﻟﻤﺘﻤﺜل ﺒﺎﻟﻁﻴﻥ ﻫﻭ ﻤﺎﺩﺓ‬
‫ﺃﻭﻟﻴﺔ ﻤﻬﻤﺔ ﻓﻲ ﺇﻨﺘﺎﺝ ﺍﻻﺴﻤﻨﺕ ﻭﻻ ﻴﻤﻜﻥ ﺇﻨﺘﺎﺝ ﺍﻻﺴﻤﻨﺕ ﺒﺩﻭﻥ ﻫﺫﻩ ﺍﻟﻤﺎﺩﺓ ‪.‬‬

‫‪123‬‬
 ‫ــــــــــــــــــــــــــــــــــــــــ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ‪...‬‬ ‫]‪[124‬‬

‫ﻜﻤﺎ ﺃﻅﻬﺭﺕ ﺍﻟﻨﺘﺎﺌﺞ ﺇﻥ ﺇﺸﺎﺭﺍﺕ ﺍﻟﻤﻌﻠﻤﺎﺕ ‪ X2‬ﻭ ‪ X3‬ﻟﻠﻨﻤﻭﺫﺝ ﺍﻷﻭل ﺴﺎﻟﺒﺔ ﻭﻫﻲ ﻤﻨﺎﻓﻴﺔ ﻟﻠﻌﻼﻗﺔ‬
‫ﺍﻟﻤﻨﻁﻘﻴﺔ ﺍﻟﺼﺤﻴﺤﺔ ﺒﻴﻥ ﻜل ﻤﻥ ﺍﻟﻤﺘﻐﻴﺭﻴﻥ ﺍﻟﺤﺠﺭ ﻭﺍﻟﻁﻴﻥ ﻤﻊ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﻤﻌﺘﻤﺩ ﻹﻨﺘﺎﺝ ﺍﻻﺴـﻤﻨﺕ‬
‫ﻭﺍﻟﺘﻲ ﻫﻲ ﻋﻼﻗﺔ ﻁﺭﺩﻴﺔ‪ ,‬ﻭﺍﻻﺨﺘﻼﻑ ﻨﻔﺴﻪ ﻅﻬﺭ ﻓﻲ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺜﺎﻨﻲ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭ ‪ X2‬ﻭﺍﻟﺘﻐﻴـﺭ‬
‫ﺍﻟﻤﻌﺘﻤﺩ ‪ Y2‬ﺇﺫ ﻅﻬﺭﺕ ﺍﻟﻌﻼﻗﺔ ﻋﻜﺴﻴﺔ ﻭﻫﺫﺍ ﻤﻨﺎﻑ ﻟﻠﻌﻼﻗﺔ ﺍﻟﺼﺤﻴﺤﺔ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴـﺭﻴﻥ ‪ ,‬ﻭﻜـﺫﻟﻙ‬
‫ﻟﻠﻨﻤﻭﺫﺝ )‪ (6‬ﺍﺫ ﻅﻬﺭﺕ ﺍﻟﻌﻼﻗﺔ ﻋﻜﺴﻴﺔ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺜﺎﻨﻲ ‪) X2‬ﺍﻟﻨﻔﻁ( ﻭﺍﻟﻤﺘﻐﻴـﺭ ﺍﻟﻤﻌﺘﻤـﺩ ‪Y3‬‬
‫)ﺍﻟﻨﻔﺎﻴﺎﺕ ﺍﻟﺼﻠﺒﺔ( ‪ ,‬ﻭﻜﺫﻟﻙ ﺍﻟﺤﺎل ﻓﻲ ﺍﻟﻨﻤﻭﺫﺝ )‪ (7‬ﺇﺫ ﻅﻬﺭﺕ ﺍﻟﻌﻼﻗﺔ ﻋﻜﺴﻴﺔ ﺍﻴﻀﹰﺎ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭ‬
‫ﺍﻷﻭل ‪) X1‬ﺍﻟﻁﺎﻗﺔ ﺍﻟﻜﻬﺭﺒﺎﺌﻴﺔ( ﻭﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﻤﻌﺘﻤﺩ ‪) Y4‬ﻜﻤﻴﺎﺕ ﺍﻟﻐﺒﺎﺭ ﺍﻟﻤﻨﺒﻌﺜﺔ( ‪.‬‬
‫ﻭﻴﻌﺯﻯ ﻭﺠﻭﺩ ﺍﻟﻌﻼﻗﺎﺕ ﻏﻴﺭ ﺍﻟﻤﻨﻁﻘﻴﺔ ﺍﻟﻰ ﻭﺠﻭﺩ ﻤﺸﻜﻠﺔ ﺘﻌﺩﺩ ﻋﻼﻗﺔ ﺨﻁﻴـﺔ ﺒـﻴﻥ ﺍﻟﻤﺘﻐﻴـﺭﺍﺕ‬
‫ﺍﻟﺘﻨﺒﺅﻴﺔ ﻭﺍﻟﺫﻱ ﺃﺩﻯ ﺒﺩﻭﺭﻩ ﺇﻟﻰ ﺘﺸﻭﻴﻪ ﻗﻴﻡ ﻤﻌﻠﻤﺎﺕ ﺍﻻﻨﺤﺩﺍﺭ ﻭﺍﺒﺘﻌﺎﺩﻫﺎ ﻋـﻥ ﺍﻟﺘﻔـﺴﻴﺭ ﺍﻟﻌﻠﻤـﻲ‬
‫ﺍﻟﺼﺤﻴﺢ‪.‬‬
‫ﻭﻟﻠﺘﺄﻜﻴﺩ ﻋﻠﻰ ﻭﺠﻭﺩ ﺘﺩﺍﺨل ﺨﻁﻲ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻓﻘﺩ ﺘﻡ ﺍﺴـﺘﺨﺩﺍﻡ ﺒﻌـﺽ ﺍﻷﺴـﺎﻟﻴﺏ‬
‫ﻟﻠﻜﺸﻑ ﻋﻨﻬﺎ ﻤﻨﻬﺎ ‪:‬‬

‫)‪variance Inflation Factors (VIF‬‬ ‫‪ – 1-1‬ﺍﺴﺘﺨﺩﺍﻡ ﻋﺎﻤل ﺘﻀﺨﻡ ﺍﻟﺘﺒﺎﻴﻥ‬

‫ﻤﻥ ﻤﻼﺤﻅﺔ ﻗﻴﻡ ﻋﻭﺍﻤل ﺘﻀﺨﻡ ﺍﻟﺘﺒﺎﻴﻥ ﻨﺠﺩ ﺃﻨﻬﺎ ﻤﺭﺘﻔﻌﺔ ﺠﺩﹰﺍ ﻭﻟﻘﺩ ﺒـﻴﻥ )‪(Marquardt,1970‬‬
‫ﻭﺠﻭﺩ ﺘﻌﺩﺩ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺨﻁﻴﺔ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻓﻲ ﺤﺎﻟﺔ ﻜﻭﻥ ﻗﻴﻡ ﻋﻭﺍﻤل ﺘـﻀﺨﻡ ﺍﻟﺘﺒـﺎﻴﻥ‬
‫‪ VIF‬ﺍﻜﺒﺭ ﻤﻥ )‪ (4‬ﺃﻭ )‪ (10‬ﻭﻤﻥ ﺨﻼل ﺍﻟﺘﺤﻠﻴل ﺍﻟﺴﺎﺒﻕ ﻴﻼﺤﻅ ﺃﻥ ﻫﺫﻩ ﺍﻟﻘﻴﻡ ﻤﺭﺘﻔﻌﺔ ﺠﺩﹰﺍ ﻭﻫﺫﺍ‬
‫ﻴﺩل ﻋﻠﻰ ﻭﺠﻭﺩ ﺘﺩﺍﺨل ﺨﻁﻲ ﻜﺒﻴﺭ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ‪.‬‬
‫‪ – 1-2‬ﺇﻴﺠﺎﺩ ﻤﺤﺩﺩ ﺍﻟﻤﺼﻔﻭﻓﺔ ‪: x ′ x‬‬
‫ﺍﻗﺘﺭﺡ ﻫﺫﺍ ﺍﻟﻤﻌﻴﺎﺭ )‪ (Mason & Webster ; 1975‬ﺇﺫ ﻴﻨﺹ ﻋﻠﻰ ﺍﻨﻪ ﺇﺫﺍ ﻜـﺎﻥ ‪x ′ x =0‬‬
‫ﻓﻬﺫﺍ ﻴﺩل ﻋﻠﻰ ﻭﺠﻭﺩ ﺘﺩﺍﺨل ﺘﺎﻡ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﺃﻤﺎ ﺇﺫﺍ ﻜﺎﻥ ‪ x ′ x =1‬ﻓﻬﺫﺍ ﻴﺩل ﻋﻠﻰ‬
‫ﺍﺴﺘﻘﻼﻟﻴﺔ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻓﻴﻤﺎ ﺒﻴﻨﻬﺎ ‪.‬‬
‫ﻭﻤﻥ ﺨﻼل ﺍﻟﻘﻴﻡ ﺍﻟﺫﺍﺘﻴﺔ ﻟﻠﻤﺼﻔﻭﻓﺔ ) ‪ ( x ′x‬ﺘﻡ ﺤﺴﺎﺏ ﻤﺤﺩﺩ ﺍﻟﻤـﺼﻔﻭﻓﺔ ) ‪ ( x ′x‬ﻭﺍﻟـﺫﻱ ﻜـﺎﻥ‬
‫ﻴﺴﺎﻭﻱ ‪:‬‬
‫‪p‬‬
‫= ‪x′ x‬‬ ‫‪∏ l j = 0.000000039‬‬
‫‪j=1‬‬

‫ﻭﻫﺫﻩ ﺍﻟﻘﻴﻤﺔ ﺼﻐﻴﺭﺓ ﺠﺩﹰﺍ ﻭﻗﺭﻴﺒﺔ ﻤﻥ ﺍﻟﺼﻔﺭ ﻭﻫﺫﺍ ﺩﻟﻴل ﺁﺨﺭ ﻋﻠﻰ ﻭﺠﻭﺩ ﺘﺩﺍﺨل ﺨﻁﻲ ﻜﺒﻴﺭ ﺒﻴﻥ‬
‫ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ‪.‬‬
 ‫]‪[125‬‬ ‫א‬ ‫‪−‬א‬ ‫א‬ ‫א‬ ‫א‬

‫‪ – 1-3‬ﺍﺴﺘﺨﺩﺍﻡ ﻋﻨﺎﺼﺭ ﻤﺼﻔﻭﻓﺔ ﺍﻻﺭﺘﺒﺎﻁ ﺍﻟﻭﺍﻗﻌﺔ ﺨﺎﺭﺝ ﺍﻟﻘﻁﺭ ‪:‬‬

‫ﺍﻗﺘﺭﺡ ﻫﺫﺍ ﺍﻷﺴﻠﻭﺏ )‪ (Gunst & Mason ; 1980‬ﺇﺫ ﻨﻼﺤﻅ ﻤﻥ ﺨﻼل ﻤـﺼﻔﻭﻓﺔ ﺍﻻﺭﺘﺒـﺎﻁ‬
‫) ‪ ( x ′x‬ﻭﺍﻟﻤﻭﻀﺤﺔ ﺃﺩﻨﺎﻩ ‪:‬‬
‫‪1.00 0.999 0.98‬‬ ‫‪0.984‬‬
‫‪0.999 1.00‬‬ ‫‪0.982 0.982‬‬
‫= ) ‪( x′x‬‬
‫‪0.984 0.982 1.00‬‬ ‫‪1.00‬‬
‫‪0.984 0.982 1.00‬‬ ‫‪1.00‬‬

‫ﻨﻼﺤﻅ ﺃﻥ ﻫﻨﺎﻙ ﺍﺭﺘﺒﺎﻁﺎ ﺸﺒﻪ ﺘﺎﻡ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ )‪ (X1,X2‬ﻭﺒـﻴﻥ )‪ (X1,X3‬ﻭﺒـﻴﻥ )‪(X1,X4‬‬
‫ﻜﺫﻟﻙ ﺒﻴﻥ )‪ (X2,X3‬ﻭ)‪ (X2,X4‬ﻭﻜﺫﻟﻙ ﺒﻴﻥ )‪ (X3,X4‬ﻤﻤﺎ ﻴﺩل ﻋﻠﻰ ﻭﺠﻭﺩ ﺘﺩﺍﺨل ﺨﻁﻲ ﻜﺒﻴﺭ‬
‫ﺒﻴﻥ ﻫﺫﻩ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﻭﺍﻟﺫﻱ ﻴﺅﺩﻱ ﺇﻟﻰ ﻋﺩﻡ ﺩﻗﺔ ﻨﺘﺎﺌﺞ ﺍﻟﺘﺤﻠﻴل ‪.‬‬
‫ﻜﺫﻟﻙ ﻨﻼﺤﻅ ﺩﻟﻴﻼ ﺁﺨﺭ ﻋﻠﻰ ﺍﻟﺘﺩﺍﺨل ﺍﻟﺨﻁﻲ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻫﻭ ﻗﻴﻤﺔ ‪ R2‬ﺍﻭ ﻤﻌﺎﻤـل‬
‫ﺍﻟﺘﺤﺩﻴﺩ ﺍﺫ ﻅﻬﺭﺕ ﻜﺒﻴﺭﺓ ﺠﺩﹰﺍ ﻤﻊ ﺍﻨﻪ ﺘﻡ ﺤﺫﻑ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺭﺍﺒﻊ ﻤﻥ ﺍﻟﻨﻤﻭﺫﺝ ﻭﻫﺫﺍ ﻤﺨﺎﻟﻑ ﻟﻠﻤﻨﻁﻕ‬
‫ﺇﺫ ﺃﻥ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺭﺍﺒﻊ ﻫﻭ ﻤﺎﺩﺓ ﺃﺴﺎﺴﻴﺔ ﻓﻲ ﺇﻨﺘﺎﺝ ﺍﻻﺴﻤﻨﺕ ‪.‬‬
‫ﻭﻟﺫﻟﻙ ﻭﻤﻥ ﺍﺠل ﺭﻓﻊ ﺍﻟﺘﺩﺍﺨل ﺍﻟﺨﻁﻲ ﻭﻤﻥ ﺜﻡ ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﻨﺘﺎﺌﺞ ﻤﻨﻁﻘﻴﺔ ﻤﻥ ﺩﻭﻥ ﺤﺫﻑ ﺃﻱ‬

‫ﻤﺘﻐﻴﺭ ﺘﻨﺒﺅﻱ ﻤﻬﻡ ﻓﻘﺩ ﺘﻡ ﺍﻟﻠﺠﻭﺀ ﺇﻟﻰ ﺃﺴﻠﻭﺏ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ )‪.(PLS‬‬

‫‪ – 2‬ﺃﺴﻠﻭﺏ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ )‪: (PLS‬‬

‫ﻜﻤﺎ ﺘﻡ ﺘﻭﻀﻴﺢ ﺃﺴﻠﻭﺏ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻓﻲ ﺍﻟﺠﺎﻨﺏ ﺍﻟﻨﻅﺭﻱ ﻭﻤﺎ ﺘﺘﻤﻴﺯ ﺒﻪ ﻤﻥ ﺤﻴﺙ‬
‫ﺇﻤﻜﺎﻨﻴﺔ ﺘﺤﻠﻴل ﺍﻟﺒﻴﺎﻨﺎﺕ ﻟﻤﺼﻔﻭﻓﺔ ﻤﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ ﻤﻊ ﻤﺼﻔﻭﻓﺔ ﻤﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴـﺔ‬
‫ﻥ ﻭﺍﺤﺩ ﻓﻘﺩ ﺘﻡ ﺇﺠﺭﺍﺀ ﻫﺫﺍ ﺍﻟﺘﺤﻠﻴل ﺒﺎﺴﺘﺨﺩﺍﻡ ﻫﺫﺍ ﺍﻷﺴﻠﻭﺏ ﻭﻜﺎﻨﺕ ﺍﻟﻨﺘﺎﺌﺞ ﻜﻤﺎ ﻴﺎﺘﻲ ‪:‬‬
‫ﻓﻲ ﺁ ٍ‬

‫ﺍﻟﺠﺩﻭل )‪ (10‬ﻤﻌﻠﻤﺎﺕ ﺍﻨﺤﺩﺍﺭ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ‬

‫‪Y1‬‬ ‫‪Y2‬‬ ‫‪Y3‬‬ ‫‪Y4‬‬

‫‪Constant‬‬ ‫‪0.016967‬‬ ‫‪- 0.013546‬‬ ‫‪- 0.013538‬‬ ‫‪- 0.806517‬‬
‫‪X1‬‬ ‫‪0.250112‬‬ ‫‪0.240652‬‬ ‫‪0.240652‬‬ ‫‪0.021208‬‬
‫‪X2‬‬ ‫‪0.249182‬‬ ‫‪0.239757‬‬ ‫‪0.239758‬‬ ‫‪0.021129‬‬
‫‪X3‬‬ ‫‪0.259901‬‬ ‫‪0.250071‬‬ ‫‪0.250071‬‬ ‫‪0.022038‬‬
‫‪X4‬‬ ‫‪0.259901‬‬ ‫‪0.250071‬‬ ‫‪0.250071‬‬ ‫‪0.022038‬‬

‫‪125‬‬
 ‫ــــــــــــــــــــــــــــــــــــــــ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ‪...‬‬ ‫]‪[126‬‬

‫ﺍﻟﺠﺩﻭل )‪ (11‬ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻹﻨﺘﺎﺝ ﺍﻻﺴﻤﻨﺕ‬

‫‪Source‬‬ ‫‪D.F‬‬ ‫‪S.S‬‬ ‫‪M.S‬‬ ‫‪F‬‬ ‫‪P‬‬

‫‪Regression‬‬ ‫‪1‬‬ ‫‪5.46298‬‬ ‫‪5.46297‬‬
‫‪363.79‬‬ ‫‪0.000‬‬
‫‪Error‬‬ ‫‪3‬‬ ‫‪0.04505‬‬ ‫‪0.01502‬‬
‫‪Total‬‬ ‫‪4‬‬ ‫‪5.50803‬‬

‫ﺍﻟﺠﺩﻭل )‪ (12‬ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻹﻨﺘﺎﺝ ﺍﻟﻜﻠﻨﻜﺭ‬

‫‪Source‬‬ ‫‪D.F‬‬ ‫‪S.S‬‬ ‫‪M.S‬‬ ‫‪F‬‬ ‫‪P‬‬

‫‪Regression‬‬ ‫‪1‬‬ ‫‪5.05755‬‬ ‫‪5.05757‬‬
‫‪363.33‬‬ ‫‪0.000‬‬
‫‪Error‬‬ ‫‪3‬‬ ‫‪0.04177‬‬ ‫‪0.01392‬‬
‫‪Total‬‬ ‫‪4‬‬ ‫‪5.09933‬‬

‫ﺍﻟﺠﺩﻭل )‪ (13‬ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻟﻜﻤﻴﺎﺕ ﺍﻟﻨﻔﺎﻴﺎﺕ ﺍﻟﺼﻠﺒﺔ‬

‫‪Source‬‬ ‫‪D.F‬‬ ‫‪S.S‬‬ ‫‪M.S‬‬ ‫‪F‬‬ ‫‪P‬‬

‫‪Regression‬‬ ‫‪1‬‬ ‫‪5.05758‬‬ ‫‪5.05759‬‬
‫‪363.07‬‬ ‫‪0.000‬‬
‫‪Error‬‬ ‫‪3‬‬ ‫‪0.04178‬‬ ‫‪0.01393‬‬
‫‪Total‬‬ ‫‪4‬‬ ‫‪5.09936‬‬

‫ﺍﻟﺠﺩﻭل )‪ (14‬ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻟﻜﻤﻴﺎﺕ ﺍﻟﻐﺒﺎﺭ ﺍﻟﻤﻨﺒﻌﺜﺔ‬

‫‪Source‬‬ ‫‪D.F‬‬ ‫‪S.S‬‬ ‫‪M.S‬‬ ‫‪F‬‬ ‫‪P‬‬

‫‪Regression‬‬ ‫‪1‬‬ ‫‪0.03928‬‬ ‫‪0.039279‬‬
‫‪0.07‬‬ ‫‪0.808‬‬
‫‪Error‬‬ ‫‪3‬‬ ‫‪1.67004‬‬ ‫‪0.556679‬‬
‫‪Total‬‬ ‫‪4‬‬ ‫‪1.70931‬‬

‫ﻴﻼﺤﻅ ﻤﻥ ﻨﺘﺎﺌﺞ ﺘﺤﻠﻴل ﺍﻨﺤﺩﺍﺭ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ ﻋﻭﺩﺓ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺘﻨﺒﺅﻱ ﺍﻟﺭﺍﺒﻊ )‪(X4‬‬
‫ﻭﺍﻟﻤﺘﻤﺜل ﺒﻤﺎﺩﺓ ﺍﻟﻁﻴﻥ ﺇﻟﻰ ﺍﻟﺘﺤﻠﻴل ﻤﻤﺎ ﻴﺸﻴﺭ ﺇﻟﻰ ﺇﺯﺍﻟﺔ ﺍﻟﺘﺩﺍﺨل ﺍﻟﺨﻁﻲ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ‬
‫ﻭﻤﻤﺎ ﻴﺅﻜﺩ ﺫﻟﻙ ﻅﻬﻭﺭ ﺍﻟﻌﻼﻗﺎﺕ ﺍﻟﻁﺭﺩﻴﺔ )ﺍﻟﻤﻨﻁﻘﻴﺔ( ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴـﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴـﺔ ﻭﺍﻟﻤﺘﻐﻴـﺭﺍﺕ‬
‫ﺍﻟﻤﻌﺘﻤﺩﺓ ﻭﺍﻟﺘﻲ ﻅﻬﺭﺕ ﻭﺍﻀﺤﺔ ﻓﻲ ﺍﻟﺠﺩﻭل )‪. (10‬‬
‫]‪[127‬‬ ‫א‬ ‫‪−‬א‬ ‫א‬ ‫א‬ ‫א‬

‫ﺍﻻﺴﺘﻨﺘﺎﺠﺎﺕ ‪:‬‬
‫‪ – 1‬ﺘﻤﻴﺯﺕ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ )‪ (PLS‬ﻋﻥ ﻁﺭﻴﻘـﺔ ﺍﻟﻤﺭﺒﻌـﺎﺕ ﺍﻟـﺼﻐﺭﻯ‬
‫ﺍﻻﻋﺘﻴﺎﺩﻴﺔ )‪ (OLS‬ﺒﺈﻤﻜﺎﻨﻴﺘﻬﺎ ﻓﻲ ﺘﺤﻠﻴل ﻤﺼﻔﻭﻓﺔ ﻤﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﺘﻤﺩﺓ ﻤﻊ ﻤـﺼﻔﻭﻓﺔ ﻤـﻥ‬
‫ﻥ ﻭﺍﺤﺩ ‪.‬‬
‫ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻓﻲ ﺁ ٍ‬

‫‪ – 2‬ﺃﻅﻬﺭﺕ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻻﻋﺘﻴﺎﺩﻴﺔ )‪ (OLS‬ﻀﻌﻔﹰﺎ ﺒﺎﻟﺘﺤﻠﻴل ﻨﺘﻴﺠﺔ ﺼﻐﺭ ﻋـﺩﺩ‬
‫ﺍﻟﻤﺸﺎﻫﺩﺍﺕ ﻋﻠﻰ ﺍﻟﺭﻏﻡ ﻤﻥ ﺇﻤﻜﺎﻨﻴﺔ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﻤﻥ ﺘﺤﻠﻴل ﺃﻴﺔ ﺒﻴﺎﻨﺎﺕ ﺒﺸﺭﻁ ﺃﻥ ﻴﻜـﻭﻥ ﻋـﺩﺩ‬
‫ﺍﻟﻤﺸﺎﻫﺩﺍﺕ ﺍﻜﺒﺭ ﻤﻥ ﻋﺩﺩ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﺒﻤﻘﺩﺍﺭ ﻭﺍﺤﺩ ‪ ,‬ﺃﻤﺎ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌـﺎﺕ ﺍﻟـﺼﻐﺭﻯ‬
‫ﺍﻟﺠﺯﺌﻴﺔ )‪ (PLS‬ﻓﺘﻤﻴﺯﺕ ﺒﺈﻤﻜﺎﻨﻴﺘﻬﺎ ﺒﺘﺤﻠﻴل ﺒﻴﺎﻨﺎﺕ ﺤﺘﻰ ﻭﺍﻥ ﻜﺎﻥ ﻋﺩﺩ ﻤﺸﺎﻫﺩﺍﺘﻬﺎ ﺍﻗل ﻤﻥ ﻋـﺩﺩ‬
‫ﻤﺘﻐﻴﺭﺍﺘﻬﺎ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻭﻫﺫﺍ ﻤﺎ ﺠﺎﺀ ﻓﻲ ﺍﻟﺠﺎﻨﺏ ﺍﻟﻨﻅﺭﻱ ﻭﺃﻜﺩﻩ ﺍﻟﺠﺎﻨﺏ ﺍﻟﻌﻤﻠﻲ ﻤﻥ ﺍﻟﺒﺤﺙ ‪.‬‬

‫‪ – 3‬ﻨﺘﻴﺠﺔ ﻟﻠﺘﺩﺍﺨل ﺍﻟﺨﻁﻲ ﺍﻟﻜﺒﻴﺭ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﻨﺒﺅﻴﺔ ﻋﺠﺯﺕ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟـﺼﻐﺭﻯ‬
‫ﺍﻻﻋﺘﻴﺎﺩﻴﺔ )‪ (OLS‬ﻋﻥ ﺇﺠﺭﺍﺀ ﺘﺤﻠﻴل ﺍﻻﻨﺤﺩﺍﺭ ﺇﻻ ﺒﻌﺩ ﺤﺫﻑ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺘﻨﺒـﺅﻱ ﺍﻟﺭﺍﺒـﻊ )‪(X4‬‬
‫ﻭﺍﻟﻤﺘﻤﺜل ﺒﻤﺎﺩﺓ ﺍﻟﻁﻴﻥ ﻭﺍﻟﺘﻲ ﻫﻲ ﻤﺎﺩﺓ ﺃﺴﺎﺴﻴﺔ ﻓﻲ ﻋﻤﻠﻴﺔ ﺇﻨﺘﺎﺝ ﺍﻻﺴﻤﻨﺕ ‪ ,‬ﻭﺒﺫﻟﻙ ﻜﺎﻨﺕ ﻨﺘـﺎﺌﺞ‬
‫ﺘﺤﻠﻴل ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﻏﻴﺭ ﻤﻨﻁﻘﻴﺔ ‪ ,‬ﻭﻜﺫﻟﻙ ﺍﻟﻌﻼﻗﺎﺕ ﺍﻟﻌﻜﺴﻴﺔ ﺒﻴﻥ )ﺍﻟﺤﺠﺭ ﻭﺍﻻﺴﻤﻨﺕ( ﻭ)ﺍﻟـﻨﻔﻁ‬
‫ﻭﺍﻻﺴﻤﻨﺕ ( ﻓﻲ ﺤﻴﻥ ﺃﻋﺎﺩﺕ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ )‪ (PLS‬ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺭﺍﺒﻊ )‪(X4‬‬
‫ﺇﻟﻰ ﺍﻟﺘﺤﻠﻴل ﻭﻅﻬﻭﺭ ﺍﻟﻌﻼﻗﺎﺕ ﺍﻟﻁﺭﺩﻴﺔ ﺒﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﻭﺍﻟﺘﻲ ﺘﻤﺜل ﻋﻼﻗﺎﺕ ﻤﻨﻁﻘﻴﺔ ﻤﻤـﺎ ﻴـﺩل‬
‫ﻋﻠﻰ ﺩﻗﺔ ﻭﻤﻨﻁﻘﻴﺔ ﺍﻟﻨﺘﺎﺌﺞ ﺍﻟﻤﺴﺘﺤﺼل ﻋﻠﻴﻬﺎ ﺒﺎﺴﺘﺨﺩﺍﻡ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﻤﻘﺎﺭﻨﺔ ﺒﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌـﺎﺕ‬
‫ﺍﻟﺼﻐﺭﻯ ﺍﻻﻋﺘﻴﺎﺩﻴﺔ )‪. (OLS‬‬

‫‪127‬‬
 ...‫ــــــــــــــــــــــــــــــــــــــــ ﺍﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ ﺍﻟﺠﺯﺌﻴﺔ‬ [128]

:‫ﺍﻟﻤﺼﺎﺩﺭ‬
1-Dante M. Pirouz ,(2006), "An Overview of Partial Least Squares "
,The Paul Merage School of Business University of California ,
Irvine, pp(1-15).
2-Frank , I. E. and Friedman ,J. H. (1993)."A Statistical View of
Chemometrcs Regression Tools". Technometrics ,35 ,pp(109-
148).
3-Gunst ,R. F. and Mason ,R. L. (1980),"Regression Analysis and its
Applications", Marcel Dekker ,Inc. New York, U.S.A.
4-Harald , M. and Luis , I. (1986)," Partial Least Squares Regression
on Design Variables As an Alternative to Analysis of Variance,
Analytica Chimica Act, 191, pp(133-148).
5-Marquardt , D. W. (1970) ,"Generalized Inverse , Ridge Regression
,Biased Linear Estimation and Nonlinear Estimation"
,Technometrics ,Vol. 12 ,pp(591-612).
6-Mason ,R. L. ,Gunst ,R. F. and Webster , I. T.,(1975),"Regression
Analysis and Problems of Multicollinearity ",Comm. In statistics
.VOL.(4) No.(3).pp(277-292).
7-Saikat Maitra and Jun Yan ,(2008),"Principle Component Analysis
and Partial Least Squares: Two Dimension Reduction
Techniques for Regression" , casualty Actuarial Society , pp(79-
90).
8-Wold , H.(1966)."La Regression PLS. Paris : Technip Iterative
Least Squares" .In P.R. Kishnaiaah (Ed). Multivariate Analysis
,New York: Academic Press .pp(391-420).