topic Re: Accounting for clustering in logistic regression model - mixed-effects logistic regression synta in Statistical Procedures

Accounting for clustering in logistic regression model - mixed-effects logistic regression syntax

jessho — Thu, 16 May 2019 12:51:45 GMT

Hi, I am trying to run a mixed-effects logistic regression model looking at the odds of receiving a prescription (rx). In my model, I am including physician and patient factors like age and sex, etc (X, Y, Z). However, I wanted to account for clustering by physician (physician_ID) in my dataset with a fixed effects model. I am having trouble figuring out which command/syntax is appropriate since this is not normal proc logit to account for clustering by physician_ID. Thank you!

Re: Accounting for clustering in logistic regression model - mixed-effects logistic regression synta

PaigeMiller — Thu, 16 May 2019 12:55:43 GMT

SInce you speak of clustering by physician, I assume you are dealing with a survey. (Is that correct?)

In that case, you should use the CLUSTER command in PROC SURVEYLOGISTIC.

Re: Accounting for clustering in logistic regression model - mixed-effects logistic regression synta

StatDave — Thu, 16 May 2019 13:56:25 GMT

If it is not survey sample data, but just clusters of correlated observations, then there are several types of models you can consider: a random effects logistic model (RANDOM statement in PROC GLIMMIX), a logistic Generalized Estimating Equations (GEE) model (REPEATED statement in PROC GEE or PROC GENMOD), or a conditional logistic model (sometimes called a "fixed effects" model, STRATA statement in PROC LOGISTIC). See the discussion and examples in the documentation for these procedures.