Herve STOLOWY <[email protected]>:
Yes, if you believe that is the right model--if the serial correlation
is the only concern, and there is no correlation of errors across i at
t (companies in a year), for example--see also
http://www.nber.org/papers/t0327. Note that -suest- is also a
cluster-robust variance estimator, so two-way clustering across two
models would imply three-way clustering, I believe. Also, the
clustering only adjusts SEs, not point estimates--you could also
include fixed effects or use a more efficient estimator (requiring a
model of the errors distributions).
On Sat, Apr 25, 2009 at 4:26 AM, Herve STOLOWY <[email protected]> wrote:
> Dear Austin:
>
> Thank you for your reply. It is helpful. I have applied the method you suggest and read the reference guide on -suest-. I have a quick question. You write:
>
> suest i v
> test [i_mean]time=[v_mean]time
> suest i v, cluster(company)
> test [i_mean]time=[v_mean]time
>
> If I understand correctly, only the second -suest- should be performed. Is that correct?
*
* 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/
