C532: ANOVA & Regression Modeling
Dec 5, 2011
Y. Cao
1
Topics:
Learn how to use residual plot to assess model assumptions for multiple
linear regression models and detect potential outliers and influential observations
by influential graphics.
Datasets
: bodyfa.sav
In a multiple linear regression, to assess the normality of the outcome variable we
may use
normal probability plot
; to see if the variances are the same for each
observation of the explanatory variable and for each treatment group then
residual
plots
in SPSS can help.
To find possible outliers and influential observations we could follow the
suggestions in today’s lecture note, for example, the scatter plot of
Cook’s D or
DfBeta
s
, DfFit
s.
If the Cook’s D is very large, we can remodel the data by excluding the outliers
and/or influential observations that has the large Cook’s D values, and compare the
models.
The influential observations can be decided (by DeBetas’ DeFits) with the statistic
values that have been given in the lecture note.
We will realize some of these in our examples.

This
** preview**
has intentionally

**sections.**

*blurred***to view the full version.**

*Sign up*