I have a regression analysis in which I have a continous outcome but I also have data which is matched on certain variables. Originally this analysis was completed by using PROC GLM (with a different sample not case-control matched) with a simple code like the following:
proc glm data=ds;
model outcome = clinical_trait_1;
weight iwm;
*also had an inverse weighting method;
run;
If I had a binary outcome I would use PROC LOGISTIC with a strata statement for a conditional logistic regression method- but in this instance I'm attempting to use the PROC GLM framework: there is no strata statement in this or PROC REG. Is there an equivalant?
A few ideas I've had so far but haven't had much success:
*PROC SURVEYREG -- Doesn't seem to have same output to match PROC REG/GLM prior to accounting for strata statement.
*PROC GENMOD with dist=normal link=identity -- Doesn't seem to match output found in PROC REG/GLM again.. Seems closer, but still differences in standard error.
*PROC GLM/PROC REG with a random statement-- would this work for a matched ID?
Any insight on this would be greatly appreciated. (on SAS 9.4)
Yes, I agree it would be nice to have strata statements in proc glm/proc reg. What it should do is to allow to include a group variable in the model, and it should be interpreted as conditioning with the sum within each level of the strata variable.
I tried apply some semiparametric theory, which in principle work like projecting the score functions of the parameters of interest into the orthogonal space to the score functions of the nuisance parameters belonging the strata variable. This give estimating equations only for the parameters of interest. (same princip as how strata works in cox regression). What a strata statement should do is then to tell the procedure to instead only solve the estimating equations for parameters in the model line, instead of the "raw" score equations.
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 16. Read more here about why you should contribute and what is in it for you!
ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.
Find more tutorials on the SAS Users YouTube channel.