st: Panel unit root, heteroskedasticity and autocorrelation

["Terzopoulou, Eleftheria" <et268@exeter.ac.uk>]

Dear Statalisters,

I am running panel with linear model(in logarithmic terms) and I found unit roots when I test with Levin Lin Chu statistic but when I check time trend and substruct the cross-sectional means it yields I(0) variables. This means that FE estimation gives correct results since it does demeaning or not? If I retrieve the combined residuals and apply the test, can it be considered as a test for cointegration?
After that I found autocorrelation applying xtserial command and also heteroskedasticity applying xttest3 command. Thus I have more problem. I found the command xtpcse that has as choices autocorr and heteroskedasticity but does it assume corr(u_i, X)= 0 as in random effects? I am in trouble because I dont know how to solve the problem. Applying Hausman's test I take the FE as preferred but the problem of autocorrelation and heteroskedastity remains.Is there any option in FE for these?

Could anyboby help me?

Thank you in advance
Elli

*
*   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/