Is there an accepted criteria for use of mvprobit over two or more
separate probit equations? My only lead so far, from the published
literature using mvprobit, tells me that mvprobit is best if outcome
variables are correlated but not so correlated that dependant variables
should be collapsed into a single variable.
