HOMEWORK NO.
BY:
EFREN JR LOCARNO LARA
PRESENTED TO:
ING JOSÉ ANGEL COLINA MÁRQUEZ
SUBJECT:
EXPERIMENTAL DESIGN
UNIVERSITY OF CARTAGENA
FACULTY OF ENGINEERING
2024
ANSWER HOMEWORK #2
Exercise 1.
To assess the impact of these two sources of variability on the random variable, we
need to compare their respective variances. This necessitates the formulation of
hypotheses. 𝐻0 = 𝜎1 = 𝜎2 , 𝐻1 = 𝜎1 ≠ 𝜎2
The exercise gave us the following data
N = 24, n = 4, a = 6, v(degrees of freedom) = N – 1 = 23
Firstly, the mean for all the released radon for all the diameters were calculated
with the general equation for the mean and then the mean of the mean that is the
summation of all the means of each hole diameter:
∑𝑎𝑖=1 𝑦𝑖
𝑦̅ =
𝑛
Then the summation of the means minus MOTM (the means of the means)
squared.
∑(𝑦𝑖 − 𝑦̅)2
The results are as follows:
𝑦̅ = 72,375
∑(𝑦𝑖 − 𝑦̅)2 = 283,34375
Then we calculate the summation of the squared of all the data minus MOTM with
the following equation:
𝑎 𝑛
2
∑ ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1 𝑖=𝑗
The result is the follow:
𝑎 𝑛
2
∑ ∑(𝑦𝑖𝑗 − 𝑦̅) = 1265,625
𝑖=1 𝑖=𝑗
For the ANOVA table the following equations were used:
• Firstable for the Treatments:
𝑎
2
𝑆𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 = 𝑛 ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1
𝑎 𝑛
2
𝑆𝑆𝑇 = ∑ ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1 𝑖=𝑗
𝑆𝑆𝑇 = 𝑆𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 + 𝑆𝑆𝐸
• Secondable for the Sum of square means:
𝑎−1
𝑁 − 1 = 𝑎𝑛 − 1
𝐷𝐹𝑇𝑜𝑡𝑎𝑙 = 𝐷𝐹𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 = 𝑎(𝑛 − 1)
𝑆𝑆
𝑀𝑆 =
𝐷𝐹
𝑠𝑠𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠
𝑀𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 =
𝑎−1
𝑆𝑆𝐸
𝑀𝑆𝐸 =
𝑎(𝑛 − 1)
𝑀𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠
𝐹0 =
𝑀𝑆𝐸
𝐹𝛼,𝑣1,𝑣2 = 𝐼𝑁𝑉. 𝐹. 𝐶𝐷 (𝛼, 𝑣1 , 𝑣2 ) ⇒ 𝐸𝑥𝑐𝑒𝑙 𝐹𝑢𝑛𝑐𝑡𝑖𝑜𝑛
𝐹0 > 𝐹𝛼,𝑣1,𝑣2
The ANOVA table is the following:
A) The ANOVA indicates that there is a statistically significant difference
between the means of the treated groups, as evidenced by the high value of
the F statistic. This suggests that the null hypothesis, which postulates that
there are no differences between the treatments, should be rejected. Most of
the total variability is explained by the differences between the groups,
confirming the effect of the treatment on the dependent variable.
B) 𝑝 − 𝑣𝑎𝑙𝑢𝑒 = 3,1595𝑥10^ − 8
Exercise 2
To assess the impact of these two sources of variability on the random variable, we
need to compare their respective variances. This necessitates the formulation of
hypotheses. 𝐻0 = 𝜎1 = 𝜎2 , 𝐻1 = 𝜎1 ≠ 𝜎2
Firstly, the mean for all the runs of each feed rate were calculated with the general
equation for the mean and then the mean of the mean that is the summation of all
the means of each hole diameter:
∑𝑎𝑖=1 𝑦𝑖
𝑦̅ =
𝑛
Then the summation of the means minus MOTM (the means of the means)
squared.
∑(𝑦𝑖 − 𝑦̅)2
The results are as follows:
𝑦̅ = 0,1045
∑(𝑦𝑖 − 𝑦̅)2 = 0,003731
Then we calculate the summation of the squared of all the data minus MOTM with
the following equation:
𝑎 𝑛
2
∑ ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1 𝑖=𝑗
The result is the follow:
𝑎 𝑛
2
∑ ∑(𝑦𝑖𝑗 − 𝑦̅) = 0,031895
𝑖=1 𝑖=𝑗
For the ANOVA table the following equations were used:
• Firstable for the Treatments:
𝑎
2
𝑆𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 = 𝑛 ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1
𝑎 𝑛
2
𝑆𝑆𝑇 = ∑ ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1 𝑖=𝑗
𝑆𝑆𝑇 = 𝑆𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 + 𝑆𝑆𝐸
• Secondable for the Sum of square means:
𝑎−1
𝑁 − 1 = 𝑎𝑛 − 1
𝐷𝐹𝑇𝑜𝑡𝑎𝑙 = 𝐷𝐹𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 = 𝑎(𝑛 − 1)
𝑆𝑆
𝑀𝑆 =
𝐷𝐹
𝑠𝑠𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠
𝑀𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 =
𝑎−1
𝑆𝑆𝐸
𝑀𝑆𝐸 =
𝑎(𝑛 − 1)
𝑀𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠
𝐹0 =
𝑀𝑆𝐸
𝐹𝛼,𝑣1,𝑣2 = 𝐼𝑁𝑉. 𝐹. 𝐶𝐷 (𝛼, 𝑣1 , 𝑣2 ) ⇒ 𝐸𝑥𝑐𝑒𝑙 𝐹𝑢𝑛𝑐𝑡𝑖𝑜𝑛
𝐹0 > 𝐹𝛼,𝑣1,𝑣2
The ANOVA table is the following:
a) The ANOVA decomposes the total variability into a component attributable to
the treatments and another to error. The results show that the sum of squares
for the treatments is significantly greater than that expected by chance. The
calculated F value falls in the rejection region, which allows us to reject the
null hypothesis of equality of means and conclude that at least one treatment
differs from the others.
b) 𝑝 − 𝑣𝑎𝑙𝑢𝑒 = 0,00233704
Exercise 3
a)
To assess the impact of these two sources of variability on the random variable, we
need to compare their respective variances. This necessitates the formulation of
hypotheses. 𝐻0 = 𝜎1 = 𝜎2 , 𝐻1 = 𝜎1 ≠ 𝜎2
Firstly, the mean for all the Outlet Flowrate for all Nozzle Design were calculated
with the general equation for the mean and then the mean of the mean that is the
summation of all the means of each hole diameter:
∑𝑎𝑖=1 𝑦𝑖
𝑦̅ =
𝑛
Then the summation of the means minus MOTM (the means of the means)
squared.
∑(𝑦𝑖 − 𝑦̅)2
Then the summation of means per block:
2
∑(𝑦𝑗 − 𝑦̅)
The results are as follows:
Outlet flowrate of the sprinkler (m/s)
Nozzle design Block1 Block2 Block3 Block4 Block5 Block6 Mean (yi-y)^2
11,73 14,37 16,59 20,43 23,46 28,74
1 0,78 0,8 0,81 0,75 0,77 0,78 0,781666667 0,00582678
2 0,85 0,85 0,92 0,86 0,81 0,83
0,853333333 2,1778E-05
3 0,93 0,92 0,95 0,89 0,89 0,83 0,901666667 0,00190678
4 1,14 0,97 0,98 0,86 0,86 0,83 0,94 0,006724
5 0,97 0,86 0,78 0,76 0,76 0,75 0,813333333 0,00199511
Mean 0,934 0,88 0,888 0,824 0,818 0,804 0,858 0,01647444
(yj-y)^2 0,005776 0,000484 0,0009 0,001156 0,0016 0,002916 0,012832
𝑦̅ = 0,858
∑(𝑦𝑖 − 𝑦̅)2 = 0,0167444
2
∑(𝑦𝑗 − 𝑦̅) = 0,012832
Then we calculate the summation of the squared of all the data minus MOTM with
the following equation:
𝑎 𝑛
2
∑ ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1 𝑖=𝑗
The result is the follow:
𝑎 𝑛
2
∑ ∑(𝑦𝑖𝑗 − 𝑦̅) = 0,22188
𝑖=1 𝑖=𝑗
For the ANOVA table the following data:
then the following equations were used:
• Firstable for the Treatments:
𝑎 𝑛
2
𝑆𝑆𝑇 = ∑ ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1 𝑖=𝑗
𝑆𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 = 𝑏 ∑(𝑦𝑖 − 𝑦̅)2
𝑖=1
𝑎
2
𝑆𝑆𝐵𝑙𝑜𝑐𝑘𝑠 = ∑(𝑦̅𝑗 − 𝑦̅)
𝑖=1
𝑆𝑆𝐸 = 𝑆𝑆𝑇 − 𝑆𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 − 𝑆𝑆𝐵𝑙𝑜𝑐𝑘𝑠
The ANOVA table was built with the following equations:
To Fcrit and p-value the following functions of Excel were used:
𝑓𝑐𝑟𝑖𝑡 = 𝐼𝑁𝑉. 𝐹. 𝐶𝐷(𝛼, 𝑣1 , 𝑣2 )
𝑝 − 𝑣𝑎𝑙𝑢𝑒 = 𝐷𝐼𝑆𝑇𝑅. 𝐹. 𝐶𝐷(𝑓0 , 𝑣1 , 𝑣2 )
Then the table looked like this:
The statistical evidence is compelling: the extremely low p-value = 0,00038167
allows us to reject the null hypothesis. We therefore conclude that there is a
significant relationship between feed rate and the standard deviation of the critical
dimension.
b) Firstly the data of each nozzle design was organice from the minimun value to
the maximun:
Nozzle Design
1 2 3 4 5
0,75 0,76 0,81 0,83 0,83
0,75 0,77 0,83 0,86 0,89
0,76 0,78 0,86 0,86 0,89
0,78 0,81 0,86 0,92 0,97
0,78 0,85 0,92 0,95 0,98
0,8 0,85 0,93 0,97 1,14
Secondly, the mean and variance were calculated using the following equation:
∑𝑁
𝑖=1 𝑥𝑖
𝑦̅ =
𝑁
∑𝑛𝑖=1
1
(𝑦𝑖 − 𝑦̅)2
𝑠2 =
𝑛1 − 1
𝑠2
𝑆𝐸 (𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑒𝑟𝑟𝑜𝑟) = √
𝑛
For the confident interval the next equations were used:
𝑦̅ ± 𝑡𝑐𝑟𝑖𝑡(𝛼⁄2, 𝑣) ∗ 𝑆𝐸
Having:
CI (Confident Interval) = 0,95
𝛼 = 1 − 𝐶𝐼
𝛼 = 1 − 0,95
𝛼 = 0,95
𝑛=4
𝑣 (𝑑𝑒𝑔𝑟𝑒𝑒𝑠 𝑜𝑓 𝑓𝑟𝑒𝑒𝑑𝑜𝑚) = 𝑛 − 1
𝛼⁄ = 0,025
2
𝑣 =5−1=4
To tcrit the Excel function 𝑉𝐴𝑅. 𝑆(𝛼⁄2 , 𝑣)
This process was repeated for each nozzle design and then the following table was
built:
Nozzle Design
alpha 0,05
1 2 3 4 5 tcrit -2,57058184
0,75 0,76 0,81 0,83 0,83 14,2857143 1
0,75 0,77 0,83 0,86 0,89 28,5714286 2
0,76 0,78 0,86 0,86 0,89 42,8571429 3
0,78 0,81 0,86 0,92 0,97 57,1428571 4
0,78 0,85 0,92 0,95 0,98 71,4285714 5
0,8 0,85 0,93 0,97 1,14 85,7142857 6
Mean 0,77 0,80333333 0,86833333 0,89833333 0,95
Variance 0,0004 0,00158667 0,00229667 0,00317667 0,0118
SE 0,00894427 0,01781385 0,02143206 0,02520582 0,04857983
Cl+ 0,74700802 0,75754137 0,81324046 0,83353971 0,82512157
Cl- 0,79299198 0,8491253 0,92342621 0,96312695 1,07487843
• In blue the variance
• In red the standard error
• In green the Limit of the Interval of the left
• In purple the Limit of the Interval on the right
Then the data of the mean was plot in a graph:
Nozzle Design vs Mean
1
0,9
0,8
0,7
0,6
Mean
0,5
0,4
0,3
0,2
0,1
0
0 1 2 3 4 5 6 7
Nozzle Design
The data of the t-distribution was plot in a graph:
T-distribution
1,2
Nozzle 1
0,8
Outlet Flowrate
Nozzle 2
0,6 Nozzle 3
Nozzle 4
0,4 Nozzle 5
Mean
0,2
0
0 20 40 60 80 100
CONFIDENT INTERVAL FOR EACH NOZZLE DESIGN
➢ Nozzle 1: 0,747008 ≤ 𝜇1 ≤ 0,792992
➢ Nozzle 2: 0,7575414 ≤ 𝜇2 ≤ 0,8491253
➢ Nozzle 3: 0,8132405 ≤ 𝜇3 ≤ 0,9234262
➢ Nozzle 4: 0,8335397 ≤ 𝜇4 ≤ 0,0,963127
➢ Nozzle 5: 0,8251216 ≤ 𝜇5 ≤ 1,0748784
Nozzle Design
1 2 3 4 5
0,75 0,76 0,81 0,83 0,83
0,75 0,77 0,83 0,86 0,89
0,76 0,78 0,86 0,86 0,89
0,78 0,81 0,86 0,92 0,97
0,78 0,85 0,92 0,95 0,98
0,8 0,85 0,93 0,97 1,14
• The nozzles that are different are 3 and 4 because not all flows are within the
confidence interval as calculated and observed in the constructed confidence
interval table.
Exercise 4
To assess the impact of these two sources of variability on the random variable, we
need to compare their respective variances. This necessitates the formulation of
hypotheses. 𝐻0 = 𝜎1 = 𝜎2 , 𝐻1 = 𝜎1 ≠ 𝜎2
Firstly, the mean for all the Stirring rate for all the furnaces were calculated with the
general equation for the mean and then the mean of the mean that is the
summation of all the means of each hole diameter:
∑𝑎𝑖=1 𝑦𝑖
𝑦̅ =
𝑛
Then the summation of the means minus MOTM (the means of the means)
squared.
∑(𝑦𝑖 − 𝑦̅)2
Then the summation of means per block:
2
∑(𝑦𝑗 − 𝑦̅)
The results are as follows:
Furnace
Stirring rate Mean (yi-y)^2
1 2 3 4
5 8 4 5 6 5,75 3,75390625
10 14 5 6 9 8,5 0,66015625
15 14 6 9 2 7,75 0,00390625
20 17 9 3 6 8,75 1,12890625
Mean 13,25 6 5,75 5,75 7,6875 5,546875
(yj-y)^2 30,9414063 2,84765625 3,75390625 3,75390625 41,296875
𝑦̅ = 7,6875
∑(𝑦𝑖 − 𝑦̅)2 = 5,546975
2
∑(𝑦𝑗 − 𝑦̅) = 41,296875
Then we calculate the summation of the squared of all the data minus MOTM with
the following equation:
𝑎 𝑛
2
∑ ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1 𝑖=𝑗
The result is the follow:
Furnace
Stirring rate
1 2 3 4
5 0,09765625 13,5976563 7,22265625 2,84765625
10 39,8476563 7,22265625 2,84765625 1,72265625
15 39,8476563 2,84765625 1,72265625 32,3476563
20 86,7226563 1,72265625 21,9726563 2,84765625
∑ 265,4375
𝑎 𝑛
2
∑ ∑(𝑦𝑖𝑗 − 𝑦̅) = 265,4375
𝑖=1 𝑖=𝑗
For the ANOVA table the following data:
Degrees of freedom
N 16
b 4
a 4
For Treatments 3
For Block 3
Total 15
Error 9
𝛼 0,05
Then the following equations were used:
• Firstable for the Treatments:
𝑎 𝑛
2
𝑆𝑆𝑇 = ∑ ∑(𝑦𝑖𝑗 − 𝑦̅)
𝑖=1 𝑖=𝑗
𝑆𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 = 𝑏 ∑(𝑦𝑖 − 𝑦̅)2
𝑖=1
𝑎
2
𝑆𝑆𝐵𝑙𝑜𝑐𝑘𝑠 = ∑(𝑦̅𝑗 − 𝑦̅)
𝑖=1
𝑆𝑆𝐸 = 𝑆𝑆𝑇 − 𝑆𝑆𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 − 𝑆𝑆𝐵𝑙𝑜𝑐𝑘𝑠
• Secondly the degrees of freedom
Degrees of freedom
a 4
b 4
N 16
Treatments 15
Blocks 3
Total 3
Error 9
The ANOVA table was built with the following equations:
𝑓𝑐𝑟𝑖𝑡 = 𝐼𝑁𝑉. 𝐹. 𝐶𝐷(𝛼, 𝑣1 , 𝑣2 )
𝑝 − 𝑣𝑎𝑙𝑢𝑒 = 𝐷𝐼𝑆𝑇𝑅. 𝐹. 𝐶𝐷(𝑓0 , 𝑣1 , 𝑣2 )
Then the table looked like this:
ANOVA TABLE
Source SS DF MS Fo Fcrit P- value
Treatments 22,1875 3 7,39583333 0,85268215 3,862548358 0,49953945
Blocks 165,1875 3 55,0625 6,34827862 3,862548358 0,01334032
Error 78,0625 9 8,67361111
Total 265,4375 23
The analysis highlights the importance of considering block effects, as they
significantly influence the model (P = 0.01334032). These findings suggest that the
variability observed between blocks is substantial. Conversely, the lack of a
significant treatment effect (P = 0.499) indicates that the treatments do not produce
discernible differences in the dependent variable.
b)
To construct a normal probability plot of the residuals, we begin by calculating the
difference between each observed value and the mean (𝑦𝑖−𝑌). These calculated
residuals are then plotted on the graph to assess their normality.
Furnace
Stirring Rate
1 2 3 4
5 2,25 -1,75 -0,75 0,25
10 5,5 -3,5 -2,5 0,5
15 6,25 -1,75 1,25 -5,75
20 8,25 0,25 -5,75 -2,75
Residuals Plot
10
8
6
4
Residuals
2
0
5 10 15 20
-2
-4
-6
-8
Stirring rate
• Analysis of the residual plot reveals that the linear regression model used
does not adequately capture the characteristics of the data. The dispersion of
the residuals is not constant, but increases as the shaking speed increases.
This 'fan'-shaped pattern suggests that the variance of the errors is not
homogeneous (there is no homoscedasticity), which contradicts one of the
fundamental assumptions of linear regression.
c)
10
4
5 rpm
Residuals
2
10 rpm
0
15 rpm
1 2 3 4
-2 20 rpm
-4
-6
-8
Furnace
• The dispersion of the residuals is irregular and depends on the stirring rate.
At low stirring rate (5 rpm), the residuals tend to be higher, while at high stirring
rate (20 rpm), they are lower.
• The assumption of homoscedasticity is significantly violated in this analysis,
manifesting in a heterogeneous variability of the residuals across the different
levels of Stirring rate. This heterogeneity questions the adequacy of the
current model and, therefore, the reliability of the statistical conclusions
obtained.
d) With increasing stirring rate, grain size usually increases. For this reason, to
achieve a reduced grain size, it is necessary to opt for a low stirring speed.
According to the data presented, the speed of 5 rpm produces the smallest
grain size, with an average of 5.75, in contrast to the higher speeds (10, 15
and 20 rpm), which result in larger grain sizes.
