Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Jen Zhen <jenzhen99@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: "cluster(): too many variables" in ivreg2 despite latest version? |
Date | Thu, 17 Oct 2013 14:00:32 +0200 |
Dear listers, (1) When specifying two-dimensional clustering in -ivreg2-, ivreg2 depvar controlvar (endogreg = instrument), cluster(dimension1 dimension2), I always get the error "cluster(): Too many variables specified". Many earlier posts on the topic on Statalist suggest that this problem should be solved by ensuring to have at least the 2010 version of -ivreg2-. While my Stata cannot be updated automatically, I have downloaded 2013 versions of both ivreg2 and of ranktest, and I have also checked -which ivreg2, all- that I currently have only that 1 version of the ado file in my PLUS ado file path. Nonetheless the problem persists. Would anyone know what else I could try? Does ivreg2 maybe require any other ado files that might also need manual updating? (2) Nichols 2007 recommends to have at least 50 different clusters and Angrist & Pischke similarly suggests to have at least 42. So I'm taking this into account when choosing on which dimensions to cluster my standard errors. What I'm wondering though is: Should I also make sure to have no or few clusters with fewer than ... observations? In particular, with the way I am currently clustering, I have a few clusters with just 2 observations, although 95% of my clusters have more than 17. With 2 way clustering, any such requirement would presumably need to be independently satisfied for each of my 2 dimensions? Thank you so much and kind regards, JZ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/