Hi everyone,

New user here. I have a logistic regression problem in which I need to compare beta weights across variables in the analysis. I am using NHANES, which of course has a complex sampling design. Rather than using PROC LOGISTIC in SAS, I had to use PROC SURVEYLOGISTIC.

I am aware that there is an STB option in SURVEYLOGISTIC, just like there is with LOGISTIC. However, is this a valid way of standardizing my parameters? I am getting very large betas.

(I know NHANES offers a tutorial on age-standardizing. That seems to be completely different from what I need... I think.)

Also, while I'm at it, I am assuming here that I cannot standardize categorical variables. How, then, can I compare the betas? What I am trying to do is say that one variable is "heavier" in predicting the outcome than another variable is.

I very much appreciate any help you can provide


Jaimie