[Limdep Nlogit List] MPROBIT with varying numbers of choices?

[Tauchmann]

Dear Limdepers,

for the Multinomial Probit model, NLOGIT allows for choice sets that vary
across individuals. Has anybody an idea how to implement "varying choice
sets" in the in the Multivariate Probit Model (MPROBIT), i.e. some
individuals have to decide about m binary choice problems while others have
to decide about subsets of m-k binary choices. Since MAXIMIZE does not allow
for multivariate normal probabilities (Mvn) it seems not to be possible to
program the likelihood function directly. Yet, as NLOGIT allows for
multivariate normal probabilities that are of different dimension across
individuals, is there a way to trick NLOGIT into estimating a Multivariate
Probit Model with varying numbers of choices?

Many thanks for any suggestions,

Best,
Harald