Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: robust standard errors

From   David De Boeck <>
Subject   st: robust standard errors
Date   Mon, 2 Aug 2010 07:21:03 -0700 (PDT)

Dear statalisters,

I have estimated a random effects model (N=84 and T=4) and I am now
inspecting the assumptions of the model. Using the FAQ guidelines, I tested
my model for heteroscedasticity (and the other assumptions). This test
confirmed that my models suffered from heteroscedasticity. I was advised to
re-estimate the model with robust standard errors. I understand that these
standard errors are robust against a heteroscedasticity bias, but I am
unsure whether the reported statictics (p-value, R_square) are valid (since
this is not the case with heteroscedasticity and the default standard errors
in xtreg re). 

Another problem is that the tests for cross sectional correlation do not
work for my dataset because either my panel is to unbalanced or I have to
few observations to complete these tests according to the errors I receive
when performing these tests (xtcsd). Is this a problem for my model? And are
there other possible tests? 

exploring the manual learned me that I can use the xtgls ,panels
(heteroscedastic) option to estimate a FGLS model for panel data in which
the assumption of heteroscedasticity is relaxed, but I am unsure how this
model relates to the random effects model (in essence my question is whether
it is also assumed that corr(u_i, X) = 0) ). Can I thus use the xtgls
function instead of the xtreg re model?

thnx in advance,
View this message in context:
Sent from the Statalist mailing list archive at
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index