Proc genmod

[jojo]

The p-value for overall estimate (type III) not significant for one variable, but the p-value for each categary of the variable are significant in proc genmod procedure. Should I use any other test instead of type III to look at the overall estimate of effect in the procedure?

[Reeza]

Sounds like a categorical variable. How did you define your levels?

[jojo]

0, 1, 2

[Reeza]

Sorry, what parameterization.

SAS/STAT(R) 9.22 User's Guide

You may want to post your code.

[jojo]

param=reference