I am using NHIS data that include survey weights in the dataset. I would like to use Proc Genmod to estimate the prevalence ratios, but I am not sure if it can account for the complex sampling design structure.
I am a Stata user, but I am sure that SAS allows you to use sampling weights in the Genmod command. I would suggest that you also investigate whether there are additional survey specific analyses in SAS that can accommodate complex survey designs. Like I said, Stata does this as a matter of course.
Yes, it can. Search for the article "Calculating Subset Weighted Analysis Using PROC SURVEYFREQ and GENMOD." The code:
proc genmod data=data;
class cluster_var stratum_var ;
model y=x /dist=bin link=log;
estimate ‘RR x1 v x0’ x 1/exp;
weight psu_var;
repeated subject =cluster_var(stratum_var);
run;
Hope that helps! Your model approach may be different. If log-binomial fails to converge, try Poisson
Did the above method (posted by Morgan) work for you?
I think you can use regular PROC GENMOD but the standard errors will be wrong. So then you can estimate the standard errors using replication methods (Balanced repeated replication).
Anyone have any further feedback? Thanks!
Hi Janet Figueroa, Thanks for sharing your insight. Could you please elaborate on how to run the Balanced repeated replication. Sharing a basic code such as that posted by Jake would be helpful. Thank you.
Now you can. Just check this last paper presented at SAS Global Forum 2017.
http://support.sas.com/resources/papers/proceedings17/0268-2017.pdf
Now you can. Just check this last paper presented at SAS Global Forum 2017.
http://support.sas.com/resources/papers/proceedings17/0268-2017.pdf
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Sampling Design