Problem in Bivariate Probit regression

[Abdul Ghafoor]

Dear Sir,

I hope this email finds you well. I am currently apply the biprobit model, the problem i am facing is that it gives an error "could not calculate numerical derivatives -- discontinuous region with missing values encountered". When i drop few variables then it gives me illogical estimation model. I am using many dummy variables in my model. I need your suggestion to control this problem. My model is given below

biprobit ( Fraud= INS Ded TRAN ETR Lev Zscore ROA PRV FinExp FOB MB BS) (Detected= Vol360 Return Turnover), partial





Thanks in advance

Regards

Abdul Ghafoor

[John Mullahy]

Abdul: One thing to check is whether all of your RHS dummy variables span both values of both outcomes. That is, for each dummy variable and each outcome does your sample contain all four combinations (0,0), (0,1), (1,0), (1,1)? An easy way to check this would be to run univariate probit separately for each outcome. The probit procedure (I recall) will drop RHS vars that have "perfect prediction" properties; I'm not sure how biprobit handles this situation but it's possible it generates the error message you report.

[Abdul Ghafoor]

Dear John: thank you so much it really helped me.