Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: pls help! correcting serial correlation in fixed effect model


From   [email protected]
To   [email protected]
Subject   Re: st: pls help! correcting serial correlation in fixed effect model
Date   Wed, 11 Apr 2007 12:02:56 +0200

The source which I suggested you says that you have unbiased standard errors if you use a panel (GLS) command and then cluster by id, so why searching for something else? However, look at the Schaffer's reply (whom I usually trust more than myself) about his command -xtivreg2-, that produces both fixed effects and newey's standard errors.
I tried with my data and I can now tell you that the following produce the same results:
xtivreg2 y $xvars, fe bw(n) robust
xi: newey2 y $xvars i.id, lag(n-1)
However, the standard errors are not exactly the same as those produced by xtreg y $xvars, fe cluster(id)
I also quote the caveats from his reply: 
> But bear in mind (a) these kernel-robust approaches [newey's] to serial
> correlation (usually) need the T in an NT panel to go off to infinity
> for the asymptotics to work, whereas most panels in practice are small-T
> large-N panels; and (b) the -cluster- option for -xtreg,fe- or
> -xtreg,re- (or -xtivreg2- for that matter) gives results that are
> consistent in the presence of arbitrary serial correlation.  -cluster-
> may be the simplest answer.

Nicola

At 02.33 11/04/2007 -0400, you wrote:
>Thank you very much. 
>Newey2 can correct serial correlation in panel data but not in fixed or random effect form. How about putting the dummies myself and then using newey2? 

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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