09-25-2014 06:50 PM
I am doing a logistic regression across a class variable "state", and the parameter always set the last state, for example, "WY"'s value as 0. I want my sample average parameters across all state to be 0. That means no state's final parameter is 0, but their average is 0. Is there a way to do it? Thanks.
09-26-2014 10:08 AM
You can use the PARAM= keyword on the CLASS statement to set the parameterization for the classification variables. Here are all the parameterizations: SAS/STAT(R) 13.1 User's Guide By default, PROC LOGISTIC uses PARAM=EFFECT. Parameter estimates for the effect coding scheme "estimate the difference in the effect of each nonreference level compared to the average effect over all levels." I am a little confused because it sounds like that is what you want, right? A comparison to the average?
Questions like this are best asked in the Statistical Procedures Community. Some of the experts there (like or ) would be able to give a more definitive answer.