Hi!

I have an ordinal variable with four level as my dependent variable. In the regression table, I have ordered probit (oprobit) estimates in the first column and generalized ordered probit (goprobit) estimates in the second, third and fourth columns. I am doing this because the parallel trends assumption does not hold and I want to compare both methods.

I have two questions:

1) How can the oprobit be interpreted relative to the goprobit models? Does it make any sense, for example, to say that it is the effect of increasing one unit of X on increasing Y in one "weighted" level? By "weighted" I mean taking into account the number of observations that responded in each level.

2) I have one variable for which none of the goprobit estimates is significant, but the one that comes from the oprobit is significant. How can I interpret this result?

Thank you in advance!