CSCI 688

Homework 4a

Megan Rose Bryant

Department of Mathematics
William and Mary

October 29, 2014

5.25 An article in the IEEE Transactions on Electron Devices (Nov. 1986, pp 1754) describes a study on
polysilion doping. The experiment shown below is a variation of their study. The response variable is base
current.
Polysilicon Anneal Temperature(Celsius)
Doping (ions) 900 950 1000
1 1020 4.60 10.15 11.01
4.40 10.20 10.58
2 1020 3.20 9.38 10.81
3.50 10.02 10.60

a.) Is there evidence with = 0.05 indicating that either polysilicon doping level or anneal temperature
affects base current?

General Linear Model: Base Current versus A, B

Method

Factor coding (-1, 0, +1)

Factor Information

Factor Type Levels Values

A Fixed 2 1, 2
B Fixed 3 1, 2, 3

Analysis of Variance

Source DF Seq SS Contribution Adj SS Adj MS F-Value P-Value

A 1 0.980 0.87% 0.980 0.9804 15.26 0.008
B 2 111.188 98.28% 111.188 55.5940 865.16 0.000
A*B 2 0.576 0.51% 0.576 0.2879 4.48 0.065
Error 6 0.386 0.34% 0.386 0.0643
Total 11 113.130 100.00%

Model Summary

S R-sq R-sq(adj) PRESS R-sq(pred)

0.253492 99.66% 99.38% 1.5422 98.64%

Since the p-values of both the polysilicon doping (factor A) and the anneal temperature (factor b) have p-
values of less than = 0.05 with values of 0.008and0.000, respectively, we can conclude that both factors are
significant at this level. Their interaction, however is not significant at the 5% level, but would be significant
at the 10% level.
b.) Prepare graphical displays to assist in interpreting this experiment.
The following graph illustrates the interaction between the factors and the mean of the base current response
variable.

c.) Analyze the residuals and comment on model adequacy.

2
The normal probability plot of the residuals is within acceptable ranges and does not cause us to question
the assumption of normality. The residuals vs predicted plot, however, causes us to question the assumption

3
of equal variances due to the odd shape. This odd grouping indicates that there might be some inequality
of variances, which would merit further examination.

d.) Is the model

y = 0 + 1 x1 + 2 x2 + 22 x22 + 12 x1 x2 + 
supported by the experiment (x1 = doping level, x2 = temperature)?. Estimate the parameters in this model
and plot the response surface.

General Linear Model: Base Current versus A, B

Analysis of Variance

Source DF Seq SS Contribution Adj SS Adj MS F-Value P-Value

A 1 0.980 0.87% 0.6262 0.6262 10.97 0.013
B 1 93.161 82.35% 18.9507 18.9507 332.00 0.000
A*B 1 0.562 0.50% 0.5618 0.5618 9.84 0.016
B*B 1 18.027 15.93% 18.0267 18.0267 315.81 0.000
Error 7 0.400 0.35% 0.3996 0.0571
Lack-of-Fit 1 0.014 0.01% 0.0140 0.0140 0.22 0.657
Pure Error 6 0.386 0.34% 0.3855 0.0643
Total 11 113.130 100.00%

Model Summary

S R-sq R-sq(adj) PRESS R-sq(pred)

0.238916 99.65% 99.44% 1.11235 99.02%

Coefficients

Term Coef SE Coef 95% CI T-Value P-Value VIF

Constant -977.5 53.0 ( -1102.8, -852.3) -18.46 0.000
A -10.64 3.21 ( -18.24, -3.04) -3.31 0.013 542.50
B 2.028 0.111 ( 1.765, 2.292) 18.22 0.000 4342.00
A*B 0.01060 0.00338 ( 0.00261, 0.01859) 3.14 0.016 551.50
B*B -0.001040 0.000059 (-0.001178, -0.000902) -17.77 0.000 4333.00

Regression Equation

Base Current = -977.5 - 10.64 A + 2.028 B + 0.01060 A*B - 0.001040 B*B

We see that since the p-value is less than our of 0.05 that all factors are deemed significant in this quadradtic
model and therefore all coefficients are significant. These coeffecients are deemed significant and calculated
above.

4
13.1 A textile mill has a large number of looms. Each loom is supposed to provide the same output of
cloth per minute. To investigate this assumption, five looms are chosen at random, and their output is noted
at different times. The following data are obtained:

Loom Ouput (lb/min)

1 14.0 14.1 14.2 14.0 14.1
2 13.9 13.8 13.9 14.0 14.0
3 14.1 14.2 14.1 14.0 13.9
4 13.6 13.8 14.0 13.9 13.7
5 13.8 13.6 13.9 13.8 14.0

a.) Explain why this is a random effects experiment. Are the looms equal in ouput? Use = 0.05.
General Linear Model: Output versus Loom

Method

Factor coding (-1, 0, +1)

Factor Information

Factor Type Levels Values

Loom Random 5 1, 2, 3, 4, 5

Analysis of Variance

Source DF Adj SS Adj MS F-Value P-Value

Loom 4 0.3416 0.08540 5.77 0.003
Error 20 0.2960 0.01480
Total 24 0.6376

Variance Components, using Adjusted SS

5
Source Variance % of Total StDev % of Total
Loom 0.01412 48.82% 0.118828 69.87%
Error 0.0148 51.18% 0.121655 71.54%
Total 0.02892 0.170059
Total 24 0.6376

This is a random effects experiment because the five looms tested were a random sampling of all of the
manufacturers looms. At the 5% significance level, the looms are not equal in output. since the p-value is
0.003.

b.) Estimate the variabilty between looms.

The variability between the looms is obtained from the above ANOVA data, but is calculated as follows.
M ST reatment M SError 0.08540 0.01480
2 = = = 0.01412
n 5

c.) Find the experimental error variance.

We obtained the experimental error variance from the previously included ANOVA data.

= M SError = 0.0148

2
d.) Find a 95 percent confidence interval for 2 +2
We know that

L 2 U
2 2
1+L + 1+U
First, we need to find L.
   
1 M ST reatment 1 1 0.08540 1
L= 1 = 1 = 0.1288
n M SError F/2,1,N 5 0.01480 3.51
Next, we must find U.
   
1 M ST reatment 1 1 0.08540 1
U= 1 = 1 = 9.6787
n M SError F1/2,1,N 5 0.01480 3.51
Now, we see that

0.1288 2 9.6787
2 2
1.1288 + 10.6787
Which reduces to the following 95% confidence interval.

2
0.1141 0.9064
2 + 2

6
e.) Analyze the residuals from this experiment. Do you think that the analysis of variance assumptions are
satisfied.

7
Based on the above graphics, we have no reason to question the normality or equal variance assumptions.
The normal probability plot has no outliers and passes the fat pencil test. The versus plots show no obvious
signs or patterns that give us cause to question our assumptions.

13.8 Refer to problem 13.1.

a.) What is the probability of accepting H0 if 2 is four times the error variance 2
We are looking for the of the test (the probability of accepting H0 . Therefore, we must first find
r r
n2 5(4 2 )
= 1+ = 1+ = 1 + 5 4 = 21 = 4.58
2 2
Now, we must determine from the operating characteristic curve. We know that

1 = a 1 = 5 1 = 4
and

2 = N a = 5 5 5 = 20
We will use an of 0.05.
Therefore, using the tables in appendix VI, we see that the probability of accepting H0 given that 2 = 4 2
is approximately 3.5%, i.e. = 0.035.

b.) If the difference between looms is large enough to increase the standard deviation of an observation by
20 percent, we wish to detect this with a probability of at least 0.80. What sample size should be used?
We know that to determine the correct sample size, we need to use the following formula:
p
= 1 + n[(1 + 0.01P )2 1]
Where
1 = a 1 = 5 1 = 4
2 = N a = 5 5 5 = 20
= 0.05
P (accepting) 0.2
Therefore
p p
= 1 + n[(1 + 0.01P )2 1] = 1 + n(1.22 1) = 1 + 0.44n
Now, we must use the operating charecteristic curves in table VI. We can see that for our , we are looking
for a [2.5, 3].

n
5 1.77
10 2.32
15 2.76

Therefore, we must select a sample size of n = 15 to detect a probability of at least 80%.

13.10 An article by Hoof and Berman (Statistical Analysis of Power Module Thermal Test Equipment
Performance, IEEE Transactions on Components, Hybrids, and Manufacturing Technology Vol 11, pp
516-520, 1988) describes an experiment conducted to investigate the capability of measurements in thermal
impedance (C/w100) on a power module for an induction moto starter. There are 10 parts, three operators,
and three replicates. The data are shown in Table 13.2.

8
a.) Analyze the data from this experiment, assuming that both parts and operators are random effects.
General Linear Model: Response versus Part, Inspector

Method

Factor coding (-1, 0, +1)

Factor Information

Factor Type Levels Values

Part Random 10 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Inspector Random 3 1, 2, 3

Analysis of Variance

Source DF Adj SS Adj MS F-Value P-Value

Part 9 3935.96 437.328 162.27 0.000
Inspector 2 39.27 19.633 7.28 0.005
Part*Inspector 18 48.51 2.695 5.27 0.000
Error 60 30.67 0.511
Total 89 4054.40

Based on the above ANOVA results, we see that every factor is significant at the 5% level since all three
factors have p-values of less than 0.05.

b.) Estimate the variance components using the analysis of variance method.
2 = M SE = 0.511
M SAB M SE
2
= n = 2.6950.511
3 = 0.728
2 M SB M SAB 437.3282.695
= an = 33 = 48.293
2 = M SA M
bn
SAB
= 19.6332.695
103 = 0.5646

13.15 Reanalyze the measurement system experiment in Problem 13.10, assuming that operators are a
fixed factor. Estimate the appropriate model components.

General Linear Model: Response versus Part, Inspector

Method

Factor coding (-1, 0, +1)

Factor Information

Factor Type Levels Values

Part Random 10 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Inspector Fixed 3 1, 2, 3

9
Analysis of Variance

Source DF Adj SS Adj MS F-Value P-Value

Part 9 3935.96 437.328 162.27 0.000
Inspector 2 39.27 19.633 7.28 0.005
Part*Inspector 18 48.51 2.695 5.27 0.000
Error 60 30.67 0.511
Total 89 4054.40

2 = M SE = 0.511
2 = M SABnM SE = 2.6950.511
3 = 0.728
2 M SB M SE 437.3280.511
= an = 33 = 48.535

10