proc surveylogistic - comparing coefficients between by variable regressions

[pamelated]

I'm currently running a surveylogistic regression with a binary by variable. I'd like to compare the coefficients beween the two models (ie when the by variable =1 versus when the by variable =0). Below is an example of the code:

proc surveylogistic data=data1;
by b;
model y (desc) = x1 x2 x3;
cluster x4;
run;

Where both b and y are binary 0/1 variables.

The output shows two separate regressions, one for when b=0 and one for when b=1. Again, I'd like to compare the coefficients across those two regressions, but can't seem to figure out how. I do not want to simply include interactions between b and all the x variables, as I'd like the regressions to be separate.

Does anyone know how to do this?

Thanks in advance!!!

[SteveDenham]

Why do the regressions separately if you wish to test for differences between the two instances?  That just doesn't make sense to me.  The interaction method should give you exactly what you need.

Steve Denham