"Joint Modeling" of Binary and Count Data in Stata

[Hm Saleh]

Hi All,

I need suggestions and commands for the "Joint Modeling" of Binary and Count Data in Stata. Briefly, I am estimating a logit model for a binary outcome, and a Poisson model for a count outcome. My data is count data.

I know SAS has a "GLIMMIX" procedure, but I need it in Stata. Thank you.

[Joseph Coveney]

Use of PROC GLIMMIX implies multilevel / hierarchical (or longitudinal) regression models. Traditionally, for mixed-response models like these, Stata users have availed themselves to a user-written command gllamm, but there are also the official Stata commands suest with a vce(cluster ...) option and for a while now gsem with different distribution families for each equation and a common random effect.

Take a look at their respective help files via

search gllamm
help suest
help gsem

for more information.

My data is count data.

This doesn't make a lot of sense. Are you saying that both outcome variables are count, but that you want to fit a binary regression model to one of them?

[Hm Saleh]

This doesn't make a lot of sense. Are you saying that both outcome variables are count, but that you want to fit a binary regression model to one of them?[/QUOTE]

Yes. I have two datasets and two outcome variables one of them is binary and the second is a count. I am estimating a logit model for a binary outcome, and a Poisson model for a count outcome.

[Hm Saleh]

This doesn't make a lot of sense. Are you saying that both outcome variables are count, but that you want to fit a binary regression model to one of them?

Yes. I have two datasets and two outcome variables one of them is binary and the second is a count. I am estimating a logit model for a binary outcome, and a Poisson model for a count outcome.