Combining CMP and Finite Mixture Models (FMM)

[Rouven Seifert]

Dear all,

I estimate a recursive system of equations using cmp command and like to adress unbserved heterogeneity within my model. Is it possible to solve this problem within the cmp command or combine the cmp with the fmm command?

Thank you!

[Weiwen Ng]

I am not familiar with the CMP command (David Roodman, available on SSC), nor with recursive models. However, the fmm prefix can't just be applied to any model. The FMM model is actually fit in the generalized structural equation modeling command (gsem). My first thought would be that gsem should be able to fit a recursive model. Unless, of course, it can't, in which case I will have to eat my words. If you can write the model syntax in gsem, then you can turn it into a mixture model with the lclass option, although this won't work if there are any continuous latent variables in the model, and it may be hard to obtain convergence.