1. In their Econometrica note in 2008, Stock and Watson suggested
using clustered standard errors with critical value of
[sqrt(n/n-1)]*t_{n-1} instead of usual t distribution for tests on
beta in models that possibly have serially correlated errors (where n
is the number of groups).
Do the t-statistics in a regression output given by stata 10
incorporate this? That is, do the results of xtreg, fe cluster(c)
provide the p-values calculated by the method of Stock and Watson or
are they based on usual t critical values? If they are based on usual
t distribution, then I should compute the Stock-Watson critical values
by hand.
2. In the same paper, S&W suggested a fix for Cov. Matrix in a FE
model. I think Stata incorporated this. Can anyone confirm?
Thank you.
*
* 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/