Could someone explain to me when it might be appropriate to use xtpcse
versus xtreg, and what the essential difference in meaning is? I have
biological cross sectional panel data, with 21 panels and 21 time points;
running it through xtpcse, xtreg, and gllamm gives similar results for the
coefficients, but the xtpcse errors are very much smaller (xtpcse with
corr(psar1) and pairwise options), and imply significance where neither
xtreg nor gllamm do.
In the animal science literature data of this type is now commonly analyzed
using multi level mixed model software like SAS Proc Mixed or S-Plus LME;
however I'd like to use Stata, which has more limited capability for
handling the multilevel models.