Quantcast
Channel: Statalist
Viewing all articles
Browse latest Browse all 72776

Biprobit marginal effects standard errors

$
0
0
Hello,

I'm estimating a recursive biprobit model for the effect of an endogenous binary variable, C, on a binary outcome, Y, with common exogenous variables, X, and excluded (instrumental) variable, Z. I'm interested in calculating the marginal effect of C. My code is as follows:

:
biprobit (Y = i.C X) (C = X Z)

margins, dydx(i.C) predict(pmarg1)
However, I get the following error message:

prediction is a function of possibly stochastic quantities other than e(b)
r(498);

After adding the "nose" option to the margins command, I'm able to get the estimate of the marginal effect, but of course without the standard errors, which are as important to me as the marginal effect. This means the problem is with the SE estimation, not the marginal effect itself

After further tinkering around, I've realized the problem only occurs if the margins command involves anything to do with the first (Y) equation, whether getting the marginal effect of any of its regressors or specifying probabilities in the "predict" options that have to do with Y.

I've searched the net, including this forum, for an explanation to this SE problem and a way to deal with it without luck. Any help would be appreciated.

P.S.: I'm using Stata 13.

Viewing all articles
Browse latest Browse all 72776

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>