Mata-based likelihood evaluators with extra variables

[Minchul Park]

I am using the ml command in Stata with mata-written likelihood evaluators.
Here, the likelihood evaluators include variables that do not appear in the equations in the ml command.
Specifically, consider the following mata-based evaluators

mata:
void my_eval(transmorphic scalar ML, real rowvector b, real colvector lnf)
{
        real vector depvar, xb, lnf
        depvar = moptimize_util_depvar(ML, 1)
        xb        = moprimize_util_depvar(ML, b, 1)

        lnf = a function including depvar, xb, and z
}
end

ml model lf my_eval() (depvar = xvar)

As you can see, the evaluator requires an extra variable z.
In this case, I am not sure how to use the input z in the ml model lf command.
How can I solve this problem?

[Joseph Coveney]

z[/B].
In this case, I am not sure how to use the input z in the ml model lf command.

Have you seen this thread? It shows one (inelegant) way.

I haven't thought this through, but you might possibly be able to accomplish this by trick, namely,specifying z as the response variable of a second empty, no-constant equation in ml model and then referring to it inside the Mata evaluator with moptimize_util_depvar(ML, 2).

Otherwise, you could as the linked thread states go all Mata and use the moptimize_init_userinfo() / moptimize_util_userinfo() function pair.

[Minchul Park]

Joseph Coveney Following your direction, I have just read the post. I think, using global with very rare name seems reasonable for now. Thank you so much.