Final - Solutions
Procedures/Theory:
- F value=1495.84; Dep Mean=103.01903; INTERCEP=47.641; x=0.372
- Normality reasonable; no outliers and no distinct bias;
homoscedasticity reasonable. Hence use
the ANOVA p-value (i.e. Pr>F: 0.2293) which is not significant. Conclude
that there is insufficient evidence to reject the null hypothesis of
equal sub-population means. Also, the conclusion is not sensitive to the
normality assumption (i.e. CHISQ p-value 0.1434).
- Remember that lower PRESS is better and assume that x1=1 indicates
subpopulation 1, x2=1 indicates subpopulation 2 etc.
- Since 1000.10 is lowest, the model is: y=beta0+beta1(x1)+error where
errors (for fixed x1) are assumed to be unbiased, homoscdastic and
normally distributed.
- Since x1=1 for subpopulation 1 and 0 otherwise then mu1=beta0+beta1 and
mu2=beta0 where mu2 is the combined mean of all other subpopulations.
SAS:
- Variables X, Y and Z exist:
- label X="Starting Salary" Y="Major" Z="GPA";
- title "Final - CSC423/324";
- proc means; var X Z;
Note: Alternatively: proc univariate; var X Z;
- Add the 'normal' and 'plot' options to proc univariate.
- Eight errors: '123' not a valid dataset name; ';' missing after 123; 'MANUFTYPE' too long; MANUFCD
and MANUFTYPE should be string; DATALINES missing; 'CLASS MANUFCD;' missing for
PROC ANOVA; PROC missing for PRINT.