Completely determined - Probit - Continuous regressor

[Marry Lee]

Hello!
I run a probit estimation and I have a regressor X that is causing problems for this estimation.

When I include this regressor, I have:

1637 failures and 740 successes completely determined.

for a sample size of 2377.

I investigated this variable X and it turns out that

• All 374 observations where the dependent variabel =1, have positive values for the regressor X, except 7 observations.
• All 1519 observations where the dependent variable =0, have negative values for the regressor X, except 22 observations.
• When I make the negative observations into positive (multiply by -1), so that all observations are positive, the problem is solved

Is there any solution to this ? please let me know if you have futher questions or if you need to see the data.

[William Lisowski]

A good place to start is by noting that you report 374 ones (successes) + 1519 zeroes (failures) = 1893 observations of your dependent variable in your exploration of the variable X, so it is not clear why probit reports at least 740 successes and 1637 failures.

[Marry Lee]

Thank you William Lisowski for pointing that out.
I made a mistake, I confused that with a subsample I made. I meant:

• All 742 observations where the dependent variabel =1, have positive values for the regressor X, except very few observations.
• All 1,646 observations where the dependent variable =0, have negative values for the regressor X, except very few observations.