Matching with replacement - Conditional Logistic Regression

SEGreaves — Tue, 22 Nov 2016 18:59:15 GMT

Hi everyone! I'm trying to figure out how PROC LOGISTIC handles repeated controls in a stratified logistic analysis, meaning that I have a 1:4 matched data set in which a couple of controls are matched to two different cases (ie. these controls show up twice in the data set with different strata values). I know matching with replacement is allowed but I'm wondering about the analysis portion, whether I can just use PROC LOGISTIC with the STRATA statement or something a little more sophisticated?

If you can point me to any readings that address this, that'd be great.

Thanks in advance for any help you can give me!

Best,

Re: Matching with replacement - Conditional Logistic Regression

Damien_Mather — Sun, 11 Dec 2016 04:03:52 GMT

I would use proc glimmix instead with the / dist=binomial model statement option and specify a random statement that partitions the variance down to an individual subject level, such as

random intercept / subject=ID;

or if your controls only appear in one (are nested in one) case-control block, and the block is identified as block,

random intercept block / subject=ID;

topic Matching with replacement - Conditional Logistic Regression in Statistical Procedures

Matching with replacement - Conditional Logistic Regression

Re: Matching with replacement - Conditional Logistic Regression