Probit/Logit: Predict failure perfectly

[Hao Ching]

Hi all,

When running the command: probit y i.x, where x is a categorical variable (x=1 if…, x=2 if… and x=3), I also obtain:

2.x !=0 predicts failure perfectly
2.x dropped and 90 observations not used

On considering the link http://www.stata.com/manuals13/rlogit.pdf, I thought this might mean that when y=1, there are only observations for x=1 and x=3, and there are no observations of x=2.

Would this be correct? Otherwise, what would STATA be referring to with this message?

Thanks in advance!

[Clyde Schechter]

No, it means that whenever x = 2, y = 1.

Parse it one step at a time:
1. 2.x, by definition, is 1 when x = 2, 0 otherwise.
2. Failure means y = 1.
3. So x = 2 predicts y = 1 perfectly.

[Joao Santos Silva]

Clyde, I believe that failure means y = 0, but I may be wrong.

Joao

[Hao Ching]

Thanks for the reply Clyde, but I'd just like to clarify if you meant failure means y=0, instead of y=1.

This is because the same link I posted states that "Stata interprets a value of 0 as a negative outcome (failure) and treats all other values (except missing) as positive outcomes (successes).", which I presume means y=0.

So that would mean that 2.x = 1 predicts y=0?

[Clyde Schechter]

Yes, sorry, I got that backwards.

[Hao Ching]

Thanks Joao and Clyde for the clarification!