Generalized residuals for control function approach

[Seungho choi]

Hi all, I am trying to correct for a model with a potential endogenous binary explanatory variable by using the control function approach. Since my endogenous variable is binary, my first stage is a probit, and the second stage is poisson or negative binominal (with the estimated errors from the first-stage as one of the explanatory variables). To implement control function approach, I need to calculate generalized residuals for the second stage. Currnetly, I made the below stata code:

In Stata:
probit y x
predict xb, xb
grhat = cond(y == 1, normalden(xb)/normal(xb), -normalden(xb)/(1-normal(xb)))

Am I using correct codes? Thank you so much for your help in advance
Best

[Alexander Huber]

Dear Seungho,

i am curious to know if you received any reply to your question or if you managed to find the answer somewhere else. Reason is, I am having the exact same question, I am lacking adequate answers however! Would be great if we could exchange some ideas how to use the CF approach and the generalized residuals. You can also reach me at [email protected]

Regards from Germany,
Alex

[Kushneel Prakash]

Any responses to this question?

[Vicente Zambrano]

Dear Seungho,

I hope this message finds you well. I am writing to inquire if you have received any responses to your question regarding the CF approach and generalized residuals. I find myself grappling with the same query and have yet to find satisfactory answers. I would greatly appreciate the opportunity to exchange insights on this matter.

Best regards from Chile,
Vicente Zambrano

[FernandoRios]

Your easiest approach is just do
predict res, score
This works for most commands