Hi all,

After adding clustered standard errors to my probit model, the F values were missing. After looking on the forum, I saw some recommendations pointing to the use of -testparm-, which I have been using. I assume the problem was that the number of variables exceeded the number of clusters (20) minus 1.

First, this is the code I have used:

Code:
probit callback i.occupation##(i.black i.woman i.parent) i.msa2, vce(cluster msa2)
testparm i.occupation##(i.black i.woman i.parent)
My questions have to do with how I should interpret the results.

Why are constraints 7, 9, 11, 17 dropped (see below)? (I know the number of constraints had to be at most 19, but why these constraints and not others?)
Can I take the result of the prob > chi2 at face value?
Is anything in this process complicated by the presence of interaction terms?
(Have I even been using the right testparm command?)

Couldn't quite make sense of it and would really appreciate some help! Thanks!

This is the stata output:


Array

(apologies for the bad quality; tried a bunch of things but couldn't get it better quality than this.)