Seemingly unrelated regression for ordered logit

[Carsten Preuss]

Dear community,

I would like to ask you whether there is a command in Stata that allows to simultaneously estimate two -ologit- regressions. The idea is to run two different dependent variables against the same set if independent variables, to see whether the effect is the same.

I know of commands like -sureg- and -biprobit- for linear and probability models, but I could not find a similar command for ordered dependent variables.

If there is no such thing, wouldn't it be possible to simply test the correlation of the errors after running both regressions independently?

Thanks a lot for your help!

Carsten

[Jenny Williams]

Try the suest command, which runs off saved model estimates for the two separate equations, or run a GSEM model with the error terms covarying. You can run a test command after both to see if the coefficients for the independent variables are equivalent.

[Carsten Preuss]

Thanks for your help, Jenny! I will try!

Best,
Carsten

[Gurdil Gill]

Hi Carsten,

Were you able to run the GSEM with correlated errors and ordered dependent variables?


Best,
Gurdil

[Maoliang Ling]

Hi Gurdil,
It seems that -gsem- cannot cope with this issue. I used -suest- and it worked. See https://www.statalist.org/forums/for...ndent-variable

[Gurdil Gill]

Hi Maoliang,

Thanks for your response.

I actually started with the suest command and it worked well for correlated errors and ordered dependent variables. However, it seems not to work when I try to introduce random effects.

I jumped to GSEM as it allowed random effects and ordered dependent variables, but does not seem to allow correlated errors.

Since you have used suest, do you happen to know how I could introduce random effects?

Gurdil