gsem command for non-recurvive model

[Maryam Bidgoli]

Hello statisticians,

I am trying to apply gsem command for non-recursive model, however, it seems gsem does not work for the non-recursive model. Any idea? I have attached a screenshot of the Stata manual.

[Dick Campbell]

I'm looking in the version 14 manual and I don't see this diagram. Please tell us which version of Stata you are using and where, exactly, you found this text. That said, it appears that Stata is telling you that you can't estimate a nonrecursive model if one of or both of the variables involved in a loop are categorical or a count (that is, a "generalized linear response"). I am not exactly sure of Stata's capabilities in this regard because I use MPlus, which will estimate the model you show.

[Dick Campbell]

I should have said I see a version of the diagram but not the surrounding text.

[Maryam Bidgoli]

Hi Dick,

Thank you so much for your quick reply.

I am working with StataMP 14.0. The text and the diagram are on the page 21 under SEM manual/ Introduction/ Intro 2.

I am estimating the relationship between health and employment status. Since there is endogeneity problem, I am estimating two equations using gsem.

When I use the following command, the result is fine.

gsem (work <- health age rabplace married seperated kid18_ lesshighschool college collegemore whitecollar2_ bluecollar wealth midwest northeast west hispanic black rsmoke rdrink) (health <- work age rabplace married seperated hispanic black midwest northeast west lesshighschool college collegemore whitecollar2_ bluecollar wealth rsmoke rdrink rconde) if (gender==1 & age<65) , cov( e.work*e.health)

However, when I add Probit and Oprobit in the end of estimations, I got the exact message error in the posted text.

[Phil Bromiley]

You might look at cmp - a user written program. I think it handles models similar to this, but whether it handles this one I'm not sure.
Phil