st: within,between and overall variation in panel data model

[wanhaiyou <wanhaiyou@hnu.edu.cn>]

Dear all,
Fixed effect model and random effect model are most used in practice.
The Hausman test can be used to choose between them. The assumption
of fixed effect model (E(alpha_i,X_it)~=0) is more intuition.However,
the variables used in the fixed effect model should be exhibit enough
variability over time. For example, the "institution quality"
variables often vary slowly over time, so that the most important information 
will be lost in the fixed effect models.My question is:
what does "enough variability over time" mean?(In stata,the xtsum can be used to
calculated the between,within and overall Std. Dev).Is there more specific rules to
make a choice? How large the within Std. Dev is needed?

Thanks for any suggestions.

Bests,

wanhaiyou

Hunan University
*
*   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/