Dear all,

I am doing some heckman estimations. However the model runds from hours without end.. I tried by estimating a probit model for the first step and then using the predicted probabilities to generate the inverse of the Mills ratio to use as a regressor in my model of interest. This approach works fine and I was wondering if the result would have been the same as if I had used the heckan command. To test this, I did both approaches in a different model and I got very similar results. However, I am wondering what are the methodological differences among both methods. Does not the heckman command do what I did by running first the probit model and then running an OLS regression in which I control for the inverste of the Mills ratio? Does anyone know the answer?

Kind regards in advance.