Hi:

I am using a two part model with a binary dependent variable and a binary endogenous regressor. I am trying to use an instrument to solve the problem of endogeneity. I thought about doing a manual 2sls where I use a linear model in the first stage and plug the fitted values in the second stage. However, I have read that it's a pretty bad idea to plug in fitted values in a non-linear model. Does anyone know what is the correct way to go about this?