Statalisters,
I estimate an ordered probit model with robust standard errors adjusted
for clustering.
LRtest seems to be not appropriate since Stata ask for "force" 'option.
How to explain it? What kind of alternatives is offered here?
I think a general rule is that, if Stata is reporting Wald chi-squares
instead of LR chi-squares, then an LR test is at least
questionable. (You'll also generally see Stata reporting log
pseudo-likelihood rather than log likelihood.) In such situations, use the
-test- command instead.
