st: What to do if panel unit root test with the demean option concludes variable is stationary

[hctang@adb.org]

Hi all,

I have ran several panel unit root tests (IPS, Fisher Dickey-Fuller and 
Phillips-Perron) using xtunitroot. I have tried several combinations of 
the available options eg., trend, drift, demean, trend and demean, demean 
and drift. Interestingly, as long as the demean option is included, the 
results tend to support the series as I(0), while other combinations are 
less conclusive. 

If I like to estimate a panel FE model, the question is whether I need to 
transform the series in any way before estimating the model, could I just 
use the series as it is (which is already in natural log) or I need to 
modify the panel FE specification in some way (if so, in what way)?

Many thanks for the advice.

Regards,
Hsiao

P.S. I am estimating a disaggregate import demand function on a panel of 
11 countries from 1991:1 to 2011:2.
*
*   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/