Dear Statalist users,

I would like to run an IV regression using different weights in the first and second stage. My reason is that while I use the same treatment variable and instrument, I assign different weights in different parts of my analysis (for good reasons) based on the outcome I have. This leads to varying F statistics in different estimations while the underlying strength of my instrument is the same. I am currently using ivreg2 which allows the use of weights but those are the same in the two stages.

The only solution I came up with is to manually estimate the two stages but retrieving the standard errors proved difficult. I would appreciate any advice on how to address the underlying issue or retrieve the correct standard errors.

Teodora