Not sure why this didn't get through the first time....
-----Original Message-----
From: Schaffer, Mark E
Sent: 04 September 2007 01:41
To: '[email protected]'
Subject: Update to -ivreg2- available from SSC
With thanks as usual to Kit Baum, an updated version of -ivreg2- by
Baum-Schaffer-Stillman is available from SSC.
To install the update from within Stata,
ssc install ivreg2, replace
Note that the new version requires the -ranktest- package by Kleibergen
and Schaffer. See the posting to Statalist
http://www.hsph.harvard.edu/cgi-bin/lwgate/STATALIST/archives/statalist.
0709/date/article-11.html, and to install -ranktest- from within Stata,
ssc install ranktest
Main changes to ivreg2:
1. Syntax simplification. In particular, the -gmm- option has been
replaced by -gmm2s-, which works in a more intuitive fashion. -gmm2s-
plus -robust-, -cluster- or -bw- will generate the corresponding
two-step feasible GMM estimates; -gmm2s- on its own is equivalent to
simple IV (since IV is just a special case of two-step GMM). The old
-gmm- option is no longer documented in the -ivreg2- help file but
-ivreg2- invoked with this option will continue behave as in earlier
versions.
2. Robust tests of underidentification and weak identification.
Previous versions of -ivreg2- reported the standard tests of
underidentification (Anderson's canonical correlations test) and weak
identification (the test of Cragg and Donald, with critical values by
Stock and Yogo). These tests are, however, valid only for the i.i.d.
case. The new version of -ivreg2- can report under- and weak
identification test statistics that are robust to heteroskedasticity,
autocorrelation, and clustering, based on the Kleibergen-Paap (2006) rk
statistic as implemented in -ranktest-.
Other additions and changes to -ivreg2- include an option for automatic
bandwidth selection, robust tests of instrument redundancy,
user-specified GMM weighting matrices, and renaming the -fwl- option to
-partial-. This last change is because -ivreg2- offers
"partialling-out" as an option for all its supported estimators, whereas
the Frisch-Waugh-Lovell Theorem applies to some but not all these
estimators. The old -fwl- option is also no longer documented in the
help file, but again -ivreg2- invoked with this option will continue to
behave as in earlier version.
More details and full references are available in the help files for
-ivreg2- and -ranktest-, and in a forthcoming discussion paper by
Baum-Schaffer-Stillman. We will announce the latter to the list when
it's available for download.
--Mark
Prof. Mark Schaffer
Director, CERT
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS tel +44-131-451-3494 / fax
+44-131-451-3296
email: [email protected]
web: http://www.sml.hw.ac.uk/ecomes
*
* 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/
