st: Autocorrelation in panel data sets (Wooldridge vs Breusch Godfrey)

["J. J. W." <bsc.j.j.w@gmail.com>]

Dear all,

I have a rather technical question to ask. I came across xtserial,
which has been programmed by David M. Drukker, which uses the
Wooldridge test procedure. However my instructor A. Kalwij claimed
that using the Breusch-Godfrey procedure was also acceptable with
panel data (it should work for fixed effects, random effects etc.).

I was wondering which test procedure was better Wooldridge or
Breusch-Godfrey? Below are both the test procedures mentioned.

The Breusch-Godfrey procedure:
- Regress the equation
- Estimate the residuals u_{i,t}
- Regress the estimated residuals on the lagged of residuals and the
other RHS variables
- Check the estimated beta and p-value of the lagged residual.

The Wooldridge procedure:
- Regress the first difference version of the model.
- Estimate the residuals of the first difference version.
- Regress the residuals with their lag counterpart
- Check whether the lagged residual is -0.5. If yes, then no autocorrelation.

PS: If I made some mistakes please correct me!

Yours sincerely,

Jun Jie Wen
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/