Using logit command and teffects ipw command to build treatment model.

[Qunyan Xu]

Hi everyone,

I used two STATA commands to build a treatment model with an aim of estimating of propensity score.

First command is

 teffects ipw

This command gives error code 322, saying that "treatment model has 2 observations completely determined; the model, as specified, is not identified". Logit was the chosen link function.

Second command is

 logit

This command runs fine. It estimates propensity scores that ranges from 0.02 to 0.97. No observations are dropped due to observations being completely determined.

The treatment model that I entered after the above two commands are identical.

Why does this happen?

Thank you for your help.

Qunyan