topic Re: Sampling Weights in logistic regression in Statistical Procedures

Sampling Weights in logistic regression

BiswadeepG — Tue, 19 Sep 2023 03:51:25 GMT

While using sampling weights in logistic regression, does SAS duplicate the rows before maximising log likelihood and estimating parameters? Is there any risk of observation independence being affected while using weights?

Re: Sampling Weights in logistic regression

StatDave — Tue, 19 Sep 2023 03:56:56 GMT

If you are analyzing survey data, do not use PROC LOGISTIC. Weights in PROC LOGISTIC do not replicate observations - they serve as multipliers in the log likelihood. Only the SURVEY procedures (SURVEYFREQ, SURVEYLOGISTIC, etc.) can provide a proper analysis of survey sample data. A variable specified in the WEIGHT statement in other procedures may produce correct parameter estimates, but their variances will not be correct. Special variance estimators are needed in the analysis of survey data and only the SURVEY procedures have these estimators. See the discussion and examples in the PROC SURVEYLOGISTIC documentation.

Re: Sampling Weights in logistic regression

BiswadeepG — Tue, 19 Sep 2023 04:04:53 GMT

Thanks a lot for your response. Is there any document reference that can help showing that the sas weights do duplicate rows ?

Re: Sampling Weights in logistic regression

sbxkoenk — Tue, 19 Sep 2023 12:23:11 GMT

Weights are NOT used to duplicate/replicate observations

, only to change the contribution of an observation to the likelihood function!

Usage Note 22600: Distinction between the WEIGHT and FREQ statements
https://support.sas.com/kb/22/600.html

Koen