How are standard errors calculated with the -xtpmg- command?

[Fredrik Bakkemo]

Hi,

I am using the -xtpmg- command by Blackburne and Frank to calculate the Mean Group (MG)-estimator. The coefficients are unweighted averages from separate time-series. However, I don't manage to reproduce the standard errors manually. Does anyone know how they are calculated? I

The program code can be found here: http://fmwww.bc.edu/repec/bocode/x/xtpmg.ado , but I don't know how to read it...


All help appreciated!

Regards,
Fredrik

[Roberto Ferrer]

That command was introduced with this reference:

The Stata Journal (2007)
7, Number 2, pp. 197-208
Estimation of nonstationary heterogeneous panels
Edward F. Blackburne III
Mark W. Frank

available here: http://www.stata-journal.com/sjpdf.h...iclenum=st0125.

The command is available from SSC and can be installed as usual:

ssc install xtpmg

[Fredrik Bakkemo]

Thank you very much! It seems I've missed something regarding the degrees of freedom. Dividing by N(N-1) (and not just N-1) yields similar standard errors. But why do they divide by N(N-1) and not (N-1)? In Blackburne and Frank (2007) they refer to Pesaran, Shin & Smith (1999), but in that article they divided by (N-1)..

---

Pesaran, H. M., Shin, Y., & Smith, R. P. (1999).
Pooled Mean Group Estimation of Dynamic Heterogeneous Panels.
Journal of the American Statistical Association, 94(446), ss. 621-634.

Blackburne, E. F., & Frank, W. M. (2007).
Estimation of nonstationary heterogeneous.
The Stata Journal, 7(2), ss. 197-208.