Re: modify Probit coefficients to match desired number of yes predictions

Thanks for your answer, m00es!

I really feel like being a blockhead with this problem.
I think your remarks brought me a good bit further but I still have
problems.

First, shouldn't Pr(y=1 | x1, x2) be the base probability of the
Probit regression rather than the complimentary probability? The
regression equation gives me for each observation the z-value and
thereby the probability that the event happens, y=1.

Second, I have the feeling that I can't leave the other predictors (x3
to x10) out when calculating the coefficients. Don't they work like a
different offset for each observation and as the standard normal
distribution is non-linear, this is important in calculating the new
coefficients?

Thanks again,
Bob

.

Quantcast