moptimize 'lf1' evaluator passed todo 2

[Richard Foltyn]

Hi,

I am implementing some MLE using moptimize, but cannot get the "lf1" evaluator method to work ("d0" and "d1" work for this problem, but I need to use SVY).

When debugging I noticed that there are numerous calls with todo=2 towards the end of the evaluation, before it stops and returns some nonsensical parameter estimates.
According to the documentation, todo = 0 or 1 for "lf1", so I do not understand what is going on. This happens both in Stata 12.1 and 13.

Thanks for any hints.
Richard

...
moptimize_init_evaluator(M, &TransMLE_lfx())
moptimize_init_evaluatortype(M, "lf1")


function TransMLE_lfx(transmorphic M, real scalar todo, real rowvector lnf, s, H) {
// here todo evaluates to 2 the last few calls during execution
}

[Jeff Pitblado (StataCorp)]

Richard found a bug in moptimize(). A fix for this will be available in a future
update to Stata 13 (and Stata 14).

While moptimize() should have passed in 1 for todo in these
cases, it is difficult to explain why this would causea problem in Richard's evaluator
without seeing the code in TransMLE_lfx().

[Richard Foltyn]

Hej Jeff,

thanks for the reply. The code in my MLE evaluator ist quite involved, that's why I did not post it. If there is any other debugging information I can provide, let me know.
I attached the value of todo by call to TransMLE_lfx, in case that is of any use.