Yue Wang
,
yuewang3,
graduate 4 credit
f) It is not appropriate to test for main effect because the interaction term is significant.
Because of hierarchy, we test for the main effect only if the interaction is not significant.
g)
Diagnostic plots are shown below:
From the Figure 1 (Residue vs. Fitted) above, it still shows a flat behavior but the
residuals for larger fitted value is apparently larger than smaller fitted values. This
could indicates heteroscedasticity of the variance. From the Figure 3 (Scale-
Location) above, we confirm that that the trend is not flat and so, we believe
there is a problem of heteroscedasticity. Figure 2 (QQ plot) is very good, thus the
normality holds.
A suggestion for fixing the hetroscedasticity problem is to do transformation to
the response.
Taking inverse is one-way. The following diagnostic plot is after
fixing.

STAT425-HW6
Yue Wang
,
yuewang3,
graduate 4 credit
Problem 2
a)
There are 4 blocks in the
eggprod
data. In the help file, we know that “six pullets