Overflow or illegal function call while forecasting probit

[mm44101]

Hi,
I am trying to calculate marginal effects for the 8 estimated coefficients in my probit model with 860 observations of which only 35 have a dependent variable of 1. The estimation of the model works fine and I get an LR of 26. However, when I try to establish fitted values by forecasting (as indicated in the manual). Eviews reports an overflow or illegal function call.

I read in other posts that this message could indicate an overfitted model and indeed, when I reduce the number of coefficients significantly, the fitting works. However, the overall significance then becomes weaker and theoretical considerations with regard to the restricted parameters are ignored.

Is there any way to solve the problem without restricting the model?

Thanks a lot. Regards, mm

[EViews Glenn]

Are you computing forecasts/fits for the same sample for which you did the estimation?