Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.

# st: RE: xtivreg2 weak-instrument-robust inference

 From "Schaffer, Mark E" To Subject st: RE: xtivreg2 weak-instrument-robust inference Date Tue, 13 Jul 2010 16:12:00 +0100

```Camden,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of
> Camden Roberts
> Sent: 13 July 2010 00:49
> To: statalist@hsphsun2.harvard.edu
> Subject: st: xtivreg2 weak-instrument-robust inference
>
> Hello all,
>
> I am attempting to use the xtivreg2 command and have a
> question about the weak-instrument-robust inference
> statistics.  In particular, when I estimate the model I
> receive the following missing statistics:
>
> Weak-instrument-robust inference
> Tests of joint significance of endogenous regressors B1 in
> main equation
> Ho: B1=0 and overidentifying restrictions are valid
> Anderson-Rubin Wald test     F(0,159)= .         P-val=     .
> Anderson-Rubin Wald test     Chi-sq(0)=.         P-val=     .
> Stock-Wright LM S statistic  Chi-sq(0)=186.00    P-val=     .
>
> With the exception of these missing statistics, the other
> tests for identification seem to suggest that the model is
> identified.  The output for these tests is below:
>
> --------------------------------------------------------------
> ----------------
> Partial R-squared of excluded instruments:   0.4127
> Test of excluded instruments:
>   F( 10,   160) =    11.24
>   Prob > F      =   0.0000
>
> Summary results for first-stage regressions
> -------------------------------------------
>
> Variable    | Shea Partial R2 |   Partial R2    |  F( 10,
> 160)    P-value
> lntotlimclas|     0.4127      |     0.4127      |       11.24
>       0.0000
>
> Underidentification tests
> Ho: matrix of reduced form coefficients has rank=K1-1
> (underidentified)
> Ha: matrix has rank=K1 (identified)
> Anderson canon. corr. N*CCEV LM statistic   Chi-sq(10)=76.75
>  P-val=0.0000
> Cragg-Donald N*CDEV Wald statistic          Chi-sq(10)=130.68
>  P-val=0.0000
>
> Weak identification test
> Ho: equation is weakly identified
> Cragg-Donald Wald F-statistic                      11.24
> See main output for Cragg-Donald weak id test critical values
>
> Underidentification test (Anderson canon. corr. LM
> statistic):          76.755
>                                                    Chi-sq(10)
> P-val =   0.0000
>
> If anyone has an idea of why I may be having this problem, I
> would greatly appreciate any assistance you could provide.

You haven't given us the regression output, but my guess would be that
you are using the cluster-robust covariance estimator, and that have as
many or more excluded instruments than you have clusters (or panel
units).

The artificial regression used for weak-ID-robust inference involves
testing all the excluded instruments.  If you have more excluded
instruments than clusters, you will run into problems because the
cluster-robust vcv won't be full rank.  In that case, the maximum number
of constraints you can test will be the number of clusters (minus 1, I
think), which isn't enough.

But this is just a guess!

--Mark

> Thank you all,
>
>
> Camden Roberts
> *
> *   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/
>

--
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.

*
*   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/
```