Hi everybody. David Roodman and Kit Baum spotted a small bug in
ivreg2 when the cluster-robust option is used - the F-stat differs
from that reported by official Stata because the degrees of freedom
adjustment is different. (We'll post a bug fix shortly.)
But this led to a question of why official Stata makes the finite-
sample adjustment it does for cluster-robust estimation, and maybe
there are some survey specialists out there who can explain it.
The manual (under Regress) states that if the option chosen is just
-robust-, the finite sample adjustment for the var-cov matrix is
N/(N-k)
where N is the number of observation and k is the number of
regressors including the constant.
If the option chosen is -cluster-, then the adjustment is
(N-1)/(N-k) * M/(M-1)
where M is the number of clusters.
The intuition is clear enough - when the number of clusters is small,
the standard errors can get big - but why M-1? Why not, for example,
M-k? (The logic for this alternative is that the rank of the var-cov
matrix is M-k when the cluster-robust option is chosen, and if M-k is
small then the standard errors should be big.) Does anybody know of
any Monte Carlo evidence on this? Inquiring minds want to know....
--Mark
Prof. Mark E. Schaffer
Director
Centre for Economic Reform and Transformation
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS UK
44-131-451-3494 direct
44-131-451-3008 fax
44-131-451-3485 CERT administrator
http://www.som.hw.ac.uk/cert
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/