Can proc genmod properly handle 'sampling' weights?

[K_S]

Hello everyone,

I have data from a survey implemented across multiple countries, so I assume my data is clustered (i.e., there is some correlation between subjects living in the same country). My outcome of interest is binary.

I was considering fitting a GEE model using PROC GENMOD, as it accounts for clustering, but I also have weights for each individual in the survey. These weights are used to correct for under- or over-representation based on factors such as age, gender, education, etc. for each country (meaning they are trying to make the sample from each country be representative of that country. 

I have reviewed the SAS documentation and understand that I can use a weight statement in PROC GENMOD, but I am unsure whether the weight statement in GENMOD functions the same way as it does in PROC SURVEYFREQ where it aims to ensures that the estimate is representative of the population . Could you provide some guidance? 

[K_S]

Hello everyone,

I have data from a survey implemented across multiple countries, so I assume my data is clustered (i.e., there is some correlation between subjects living in the same country). My outcome of interest is binary.

I was considering fitting a GEE model using PROC GENMOD, as it accounts for clustering, but I also have weights for each individual in the survey. These weights are used to correct for under- or over-representation based on factors such as age, gender, education, etc. for each country (meaning they are trying to make the sample from each country be representative of that country. 

I have reviewed the SAS documentation and understand that I can use a weight statement in PROC GENMOD, but I am unsure whether the weight statement in GENMOD functions the same way as it does in PROC SURVEYFREQ where it aims to ensures that the estimate is representative of the population . Could you provide some guidance? 

[Kurt_Bremser]

Please do not double-post.

[K_S]

I won't do it again. I wasn't aware this counts as double posting. I thought I was posting in two separate forums with different foci and groups of experts commenting. 

[Kurt_Bremser]

GENMOD is a part of SAS/STAT, so you're good right here. If you accidentally post something in what you feel is the wrong community, notify one of use superusers, and we'll move it for you.

But multiple questions for the same issue only make it harder to track the path to a possible solution.

[StatDave]

If you are analyzing survey data, 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. There is currently no SURVEY procedure for Generalized Estimating Equations models.

[K_S]

Is it possible to account for  the within country correlation in PROC SURVEYLOGISTIC? 

[Season]

Yes, it can.