Heckit starting values

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

Post Reply
startz
Non-normality and collinearity are NOT problems!
Posts: 3798
Joined: Wed Sep 17, 2008 2:25 pm

Heckit starting values

Postby startz » Wed Apr 29, 2015 9:29 pm

How does EViews pick starting values for the correlation coefficient when doing a Heckit by mle?
Top
EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 13604
Joined: Tue Sep 16, 2008 5:38 pm

Re: Heckit starting values

Postby EViews Gareth » Thu Apr 30, 2015 7:48 am

It uses the estimates calculated as part of the two-step procedure. Essentially ML estimates the full two-step first, and uses those as starting values. Equation 28.43 in EV9 UGII.
Top
startz
Non-normality and collinearity are NOT problems!
Posts: 3798
Joined: Wed Sep 17, 2008 2:25 pm

Re: Heckit starting values

Postby startz » Thu Apr 30, 2015 7:51 am

I see where rho*sigma comes from. Is sigma done from y-x*betaHat from the two-step?
Top
EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 13604
Joined: Tue Sep 16, 2008 5:38 pm

Re: Heckit starting values

Postby EViews Gareth » Thu Apr 30, 2015 8:07 am

You can calculate sigma from the SSR of the regression of the response equation, adjusting it for the mills-ratio terms.

sigma = sqrt(SSR/N + theta*theta*mean(lambda))
Top
startz
Non-normality and collinearity are NOT problems!
Posts: 3798
Joined: Wed Sep 17, 2008 2:25 pm

Re: Heckit starting values

Postby startz » Thu Apr 30, 2015 8:08 am

Got it. Thanks.
Top
Post Reply

Return to “Estimation”

Who is online

Users browsing this forum: No registered users and 2 guests