Page 2 of 2
Re: Coputation of fitted values
Posted: Thu Sep 21, 2017 12:20 pm
by oksanakim
But this only confirms what I mentioned earlier that Heckit and manual estimation return completely different results. :D
Re: Coputation of fitted values
Posted: Thu Sep 21, 2017 12:37 pm
by EViews Gareth
Again, without knowing exactly what you're doing, there's no way for us to help.
To start, if I were you, I'd take a look at what the code I posted does, take a look at what you're doing, and then compare to see how they differ.
Re: Coputation of fitted values
Posted: Thu Sep 21, 2017 12:40 pm
by oksanakim
I estimate the first stage Probit (selection stage) (without a code, built-in options in Eviews) and then compute the Inverse mills ratio using the code. Then, I estimate OLS (built-in, not a code) and explicitly include a previously stored Imills as an additional explanatory variable. I don't think there is anything wrong with this algorithm, which is more transparent than Heckit. Plus, you confirmed my code for Imills calculation was valid. Thanks anyway.