Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

st: RE: RESET with fixed effects

From   "Schaffer, Mark E" <[email protected]>
To   <[email protected]>
Subject   st: RE: RESET with fixed effects
Date   Wed, 14 Sep 2011 18:48:06 +0100


I would stick with my original suggestion in the Statalist post you are
pointing to, i.e., apply the within-transformation to the data and then
do a RESET test in the usual way.

One rationale is the one you are pointing to (which would apply for any
large-N asymptotics) - the estimated FEs are inconsistent.

The other is the interpretation I've just offered in response to your
other question.  The FE estimator is a kind of GLS estimator, and you're
working with GLS-transformed data.  If you're not sure about this
particular GLS transform (i.e., the within-transformation), though, you
could investigate that first.


> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of Bruno Rocha
> Sent: 14 September 2011 17:32
> To: statalist
> Subject: st: RESET with fixed effects
> Dear statalisters, hello,
> I have implemented a panel FE model and I would like to 
> implement a RESET test to see if the linear specification of 
> the model is correct.
> However I don't know whether I should include the country and 
> year fixed effects in the fitted values computed from the 
> original equation. Here 
> it is suggested that data should be demeaned and then perform 
> the test.
> Would it be wrong to compute the squared fitted values using 
> the FE? A possible objection arises from the fact that the 
> coefficients of country dummies are in principle inconsistent 
> (as T is only 8 and N=27).
> Many thanks for your attention,
> Best regards
> Bruno Rocha
> (Phd student -- University of Essex)
> *
> *   For searches and help try:
> *
> *
> *

Heriot-Watt University is a Scottish charity
registered under charity number SC000278.

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index