Re: st: Reemedy for serial correlation in Panel Data

[David Greenberg <dg4@nyu.edu>]

Small sizes are going to mean large standard errors, or to put it
differently, low power to detect small to medium-sized effects.
Moreover, some statistics for panel data are only valid
asymptotically, with distributions in small samples that are not
known. Whether it makes sense for you to proceed in these
circumstances is hard to say without knowing more about what you are
doing. David Greenberg

On Mon, Apr 9, 2012 at 10:48 PM, D. Demetriou <dd358@cam.ac.uk> wrote:
> Dear all,
> Thank you for your valuable comments!
>
> David (Greenberg) do you believe that the size of my data(T=10, N=9) is
> forbidding for empirical research?
>
> I am using Regional Trends which unfortunately goes back 10 years the most
> for my variables of interest. And regions ofcourse are limited to 9. Data on
> subregions (so that I could increase N) does not exist and where it exists
> goes only back to 2-3 years.
>
> I have taken all precautions to maximise the reliability of results and
> intensively analyse residuals by region and panel after each regression to
> ensure that coefficients are not the outome of 1 or few influential
> observations. I also apply all sort of tests: collinearity, normality of
> idiosyncratic component(to enable correct inference from t-statistics),
> cross sectional dependance, etc
>
> So, do you think there is no way an empirical research should move forward
> with such a small size?
>
> PS: It will come a bit hard on me as I am halfway the paper!
>
> Yours,
>
> DD
> *
> *   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/

*
*   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/