Proc genmod error

[SteveDenham]

Looking at the discussion on the MedStats Google Group, I see where the recommendation has grown to fitting a cubic spline through the logits.  If you decide to go this way, look into the EFFECT statement.

Steve Denham

[Pooja]

Will do so, Steve.

Thank you so much for your help!

Pooja

[SureshK]

Pooja,

Based on the SAS Error, I looked at your model statement and noticed a lot of variable names (18 of them). Many of them are not in the class statement!!! May be you excluded them in the class statement since those variables are not class variables. Check in to that.

Posting first few lines of sample data will also help diagnosing the issue.

Also, decide on the most important (2 - 3) independent variables, co-variance structures and distributions first and try to get the output. Then you can build on the complicated model by adding in one variable at a time.

Suresh