st: Question on panel data estimation

[Narasimhan Sowmyanarayanan <narasimhan.sowmyanarayanan@gmail.com>]

I am working on an unbalanced dynamic panel. I am trying to use the
xtdpdsys command since I want to do the estimation in level and the
robust option to get heteroskedasticity consistent estimator to serial
correlation. First I used the one step estimator and the standard
error from the robust option in the one step estimator is not too
different from the original standard error without the whites
correction. However, when I use the two step estimator, with the
robust option, the standard errors drastically increase and all the
significant effect that I obtained from the one step estimator with
the robust standard error option disappear. For example the z value of
a robust one step estimator is around 4 while the uncorrected two step
estimator has a z-value of 100 and the Windmeijer corrected two step
estimator (with the robust option) gets me a z value of 1. I want to
understand if these wild swings are a nature of the data or is it
something to do with the estimation. Some insights would be highly
appreciated.

Thanks in advance

Narasimhan.
*
*   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/