Re: st: correction for within panel serial correlation in standard errors

[Stas Kolenikov <skolenik@gmail.com>]

On Fri, Apr 20, 2012 at 2:49 PM, Yogesh Uppal
<stata.list.queries@gmail.com> wrote:
> Dear All,
>
> I want to check sensitivity of my results with respect to possible
> within-panel serial correlation in a sample for 50 US states over 28
> years.
>
> I am trying following methods to compute the standard errors: cluster,
> xtscc, bootstrapping.
>
> I have a few questions regarding each of these:
>
> 1) How do I determine which of these methods is more suited to my data?
>
> 2) While bootstrapping SEs, is there a method to determine the optimal
> number of repetitions?

Your bootstrap must respect the existing correlations structure,
otherwise, your standard errors will be too small. I talked about
survey bootstraps with Stata in
http://stata-journal.com/article.html?article=st0187.

If the various methods give you wildly different answers, you should
not trust any. Correctly implemented, the cluster standard errors and
the bootstrap standard errors should be about the same.

-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
*   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/