Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

AW: st: RE: ordered probit panel data model with fixed effects


From   "Martin Weiss" <[email protected]>
To   <[email protected]>
Subject   AW: st: RE: ordered probit panel data model with fixed effects
Date   Thu, 20 Aug 2009 21:39:53 +0200

<> 

To echo Nick`s sentiment, Stata is extermely conservative with regard to
warnings. There were calls for more comprehensive and frequent ones at one
of the American UGMs last year, but luckily the policy pursued by StataCorp
does not seem to have changed in Stata 11. 

Particularly given that you have many covariates in your model (which rarely
washes with the referee anyway...) you should be glad that Stata is so kind
to provide you with this warning, as this condition would be impossible to
spot otherwise.

Also note that the Reference manual [R] devotes all of 6 pages (904-910) to
the subject. So ignoring this message should not even make it onto your list
of options...



HTH
Martin

-----Ursprüngliche Nachricht-----
Von: [email protected]
[mailto:[email protected]] Im Auftrag von Nick Cox
Gesendet: Donnerstag, 20. August 2009 21:30
An: [email protected]
Betreff: RE: st: RE: ordered probit panel data model with fixed effects

What kind of answer do you expect to this kind of question? 

You have a clear signal from Stata that what you asked for makes limited
-- perhaps little or no -- statistical sense. 

Don't ignore it. 

Think what it means. In particular, are you trying to fit a model that
is far too complicated? 

I don't have specific advice for you -- just general advice that only
rarely is it a good idea to ignore Stata warning messages. 

Nick 
[email protected] 

J. Li

Thanks so much for your reference to the webpage. I looked into the
page you mentioned. It talks about two different cases which may cause
this problem 1) this happens when a continuous variable (or a
combination of a continuous variable with dummy variables) is a great
predictor; and 2) hidden collinearity-- this happens when all the
regressors are dummy variables. Since my regressor list includes both
continuous variables and dummy variables (a lot of individual dummies),
can I simply ignore this warning message of "4,305 observations
completely determined. s.e. questionable" after my regression? Thanks!

On Fri, 14 Aug 2009 20:56:59 +0200
 "Martin Weiss" <[email protected]> wrote:
 
> http://www.stata.com/support/faqs/stat/logitcd.html

J. Li
 
> I have a question about running ordered probit panel data model with
> fixed effects. I know the discussion on the incidental parameters
> problem with using fixed effects nonlinear panel data model with
> large
> N and small T. However, my work involves estimating the basic ordered
> probit panel data model with individual fixed effects before applying
> adjustment to the estimates so that the bias could be corrected with
> respect to the parameters of interests. My problem now is with the
> simple estimation with ordered probit fixed effects model. The
> estimation line can be simplied as this: 
> 
> xi: oprobit health education income age gender i.ID, robust
> cluster(individual)
> 
> where i.ID are all the individual dummies (there are 3,733 dummies in
> my case)
> 
> After the estimation, I got the warning message as: 4,305
> observations
> completely determined. s.e. questionable. (note that this panel is
> with
> N=3,733 and T=6). Should I be concerned with this warning message?
> What
> does this mean? And what should I do to solve this problem? Thank you
> so much!

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index