heckprobit postestimation - why the predicted probabilities return missing values for who does not participate?

[Maria Luiza Garcia Rosa]

I am working with probit model with selection (heckman correction) for correcting estimates of heart failure prevalence from a sample survey. The problem is that the predicted probabilities p margin, p11, p10, p01 p00 and pcon , estimated with the comands of heckprobit postestimation returns missing values for those not participating in the study. Just psel has actual values for the entire sample. I tried to figure out why and how to correct the two models, without success. Could I have some help? Thanks

[Stephen Jenkins]

We need to see exactly what you typed into Stata and exactly what Stata returned. (Please see the Forum FAQ on this; especially section 12 and how to use CODE delimiters to report such information.)
If I look at help heckprobit postestimation##predict, I see that there are many potential prediction variables, some of which condition on selection and some don't. Also show us the results of a describe and summarize command for the 2 outcome variables and the predictors, and the predicted variables.

[Maria Luiza Garcia Rosa]

Hi Stephen,

Thanks a lot for yours comments. I was making a basic error and I fixed it. But now I have another question in the same subjetct: heckprob. One of the assumptions of biprobit Heckman selection is: the conditional joint error distribution must be bivariate Normals . Cant we test this in stata?
Thanks once again and sorry about my bad english.