Hi,
I am running IV regression with clustered robust standard errors. I am
getting the following error message :
estimated covariance matrix of moment conditions not of full rank.
standard errors and model tests should be interpreted with caution.
Possible causes:
number of clusters insufficient to calculate robust covariance matrix
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
I figured that this problem can be addressed by using the "partial()"
option, to remove enough exogenous regressors for S to have full rank.
I am running the following regressions-
xi: ivreg2 no_bachg black_pct crime_rt pop_mile i.YEAR i.MSA
(ratio=JAN_MIN),first cluster()
xi: ivreg2 no_bachg black_pct crime_rt pop_mile i.YEAR i.MSA
(ratio=JUL_MAX),first cluster()
xi: ivreg2 no_bachg black_pct crime_rt pop_mile i.YEAR i.MSA
(ratio=JAN_MIN JUL_MAX PREC),first , cluster()
I am clustering by county, MSA and also county_year and msa_year.
I want to know whether there is any criterion for choosing the
variables used for partitioning. In my case the model can predict
robust clustered standard errors (i.e. without the above error
message), only when I am partitioning all the variables, except the
endogenous and excluded instruments and clustering by msa_year. I have
not checked whether the model can predict the robust standard errors
if I cluster by either county or MSA.
Also what is the "fwl()" option in this context and how is it
different from "partial()".
The other thing is that using STATA's "ivreg" for the same set of
regressions I do not get the above error messages, so is it then
better for me to use "ivreg" instead of "ivreg2" ?
I would really appreciate your feedback on this.
Thank you.
Rita.
*
* 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/
