How to control the significance level for the final set of variables in PROC LOGISTIC

[Ruth]

Hi,

I used PROC LOGISTIC to run a model using backward variable selection method. For this method, the final set of variables that were kept in the model have p-values >0.05 for the Type3 analysis. This is not good if I want to keep only those variables with p-value<=0.05. Can I ask how I can control the significance level for the final variables in PROC LOGISTIC?

THANKS.

[art297]

Will setting the alpha= option do what you want?  see, e.g.: http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_logistic_sec...

[Ruth]

It is SLSTAY/SLENTRY.

[StatDave]

If you have interactions in the model, the default is to maintain a heirarchical model.  That is, if A*B is in the model, then A and B are required to be in the model.  In such a case, A or B may not meet the significance level requirements.  You can turn off model hierarchy with the HIER= option in the MODEL statement.