Let me supply my own answer for the benefit of the archives and future questioners:
(1) xtreg does allow you to "predict" residuals, but for some reason the "option" to use is "e" rather than the "r" used after reg.
Examples: (a) reg dv iv; predict residuals, r; (b) xtreg dv iv, predict residuals, e.
(2) customer support has confirmed that stata 9.2 for some strange reason doesn't support any automatic calculations of residuals after xtpcse; as I should have been able to figure out on my own, it is trivially easy to compute these mechanically by predicting xb and then subtracting xb from the dependent variable.
Example: xtpcse dv iv; predict xb if e(sample); gen residuals = dv-xb
It would be nice if Statacorp incorporated more of the reg postestimation capabilities into the xt series of estimation commands. xtpcse seems particularly underserved.
JY
________________________________
From: [email protected] on behalf of Jason Yackee
Sent: Tue 8/22/2006 8:18 PM
To: [email protected]
Subject: predicting residuals with xtpcse
Dear Statalisters:
I fear this question may demonstrate that there is such a thing as a stupid one, but several hours of head-scratching my way through my stata manuals, the archives here, Baltagi, and so on have failed to part the clouds from my thinking.
So in short: why can I not use the "predict residuals, r" routine to generate model residuals for xt--- estimations like xtpcse? From the manuals I understand that this is not possible. Is there some other way I can generate my residuals in order to plot them against fitted values? If not, is there an econometric reason why this is not possible when using xt estimation strategies with panel data?
Thank for any advice you can provide. If it matters (and I can't imagine it does), I am using Versions 8.2 & 9.2, Intercooled.
JY
<<winmail.dat>>
