Forum: help

Model type of second stage with binary outcome [ Reply ]
By: Sebastian Hellmeier on 2020-04-15 10:49

Hi everyone,

I am running a selection model with a binary outcome variable. My goal is to simulate predicted values based on the model output over the range of my main independent variable of interest.

For the selection stage, this is fairly straightforward as the selection stage is a probit model. My question is which distribution/model is used for the second stage if the outcome is binary?

Maybe I am misunderstanding how the second stage is calculated but it was not clear to me which type of model is used (logit/probit). For example, when I run predict(model, type = "unconditional", part = "outcome"), I get only negative values and I am not sure how to interpret them.

Any advice is appreciated. Many thanks in advance.

Best,
Sebastian