Dear Stata Users,

I have a sample of 14,310 observations, consisting of trade data between 54 different countries over 5 years. I made it a balanced panel by using zero as the trade value for observations that previously did not exist in order to be able to use a truncated regression. However, as I tried running the following command:

Code:
truncreg $ylist $xlist, ll(0)
margins, dydx(*) atmeans predict (e(0,.))
I got the following result:

(note: 7060 obs. truncated)

Fitting full model:

Iteration 0: log likelihood = 3833.8207
Iteration 1: log likelihood = 7625.6716
Iteration 2: log likelihood = 19343.622
Iteration 3: log likelihood = 20343.974
Iteration 4: log likelihood = 20557.564
Iteration 5: log likelihood = 20650.613
Iteration 6: log likelihood = 20718.359
Iteration 7: log likelihood = 20726.809 (backed up)
Iteration 8: log likelihood = 20729.282 (backed up)
Iteration 9: log likelihood = 20730.646 (backed up)
Iteration 10: log likelihood = 20730.832 (backed up)
Iteration 11: log likelihood = 20730.842 (backed up)
numerical derivatives are approximate
nearby values are missing
Iteration 12: log likelihood = 20730.85 (backed up)
numerical derivatives are approximate
nearby values are missing
cannot compute an improvement -- discontinuous region encountered
Not only was convergence not achieved, but also the log likelihood started increasing with each iteration instead of decreasing. I already tried rescalling the dependent variable, but it did not work.

Can anyone help me? Thank you in advance.

Note: I am using Stata/MP 12.1 for Windows (64-bit x86-64) on a PC with 4GB RAM.

Kind regards,
Pedro