Re: [Gretl-users] Nested Logit references on nonlinear panel models (fixed effects probit included)
Am 07.12.2014 um 00:13 schrieb oleg_komashko(a)ukr.net:
non-technical survey
http://methods.johndavidpoe.com/2014/02/07/fixed-effects-probit/
To be honest, some of the negative remarks on that page sound like my
own thoughts before I had really learned and understood the theory. I
would not recommend to follow that advice. Also, if you take it
seriously that fixed effects are bad because you are left only with the
within information, you're (=the author of the page) missing the point I
think.
Methods for bias reduction
1. Jinyong Hahn and
2. Whitney Newey
3.
Jackknife and Analytical Bias Reduction for Nonlinear Panel Models
Eco nometrica Volume 72, Issue
4,<http://onlinelibrary.wiley.com/doi/10.1111/ecta.2004.72.issue-4/iss...
1295–1319,July 2004
When I wrote "don't know" I only meant that I wasn't sure about exactly
what's available in gretl or in function packages. Of course there is a
large literature out there for all kinds of problems and how to do
specialized things. Anybody is welcome to implement these things (as
function packages) but it's not clear (to me) why it should be a
priority, because it's of course only a partial solution to the probit
fixed effects bias.
cheers,
sven