Statalist


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

Re: st: Autocorrelation tests for FE models


From   Austin Nichols <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Autocorrelation tests for FE models
Date   Thu, 4 Jun 2009 21:29:42 -0400

In the presence of this "too high autocorrelation" that is
statistically nonzero, you should cluster by group.

On Thu, Jun 4, 2009 at 7:57 PM, Kyle Hood <kyle.hood@yale.edu> wrote:
> There is no conflict, here.  Stata has not reported any standard errors for
> rho estimated using the xtregar command; this command is not testing for
> serial autocorrelation, per se, although if you can find a standard error
> for rho in the e() structure, maybe you can do the test yourself.  The other
> command you report the results for is testing for serial autocorrelation,
> and we only know the p-value associated with this test.  Be careful not to
> confuse statistical significance and effect size -- both of these commands
> are telling you different things.  The latter test strengthens the argument
> that xtregar should be used (instead of xtreg, for example).
>
> P K wrote:
>>
>> Hi,
>>
>> I test a fixed effects panel model with 61 groups for 10 years and tested
>> for autocorrelation with the xtregar and the xtserial command, which lead to
>> different findings. Xtregar doesn't indicate too high autocorrelation
>> (rho_ar =.12), while xtserial indicates autocorrelation (Woolridge test:
>> Prob > F =      0.0041)
>>
>> My questions therefore are:

*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index