Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
| From | "David Roodman (droodman@cgdev.org)" <DRoodman@cgdev.org> |
| To | "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |
| Subject | st: new xtabond2 feature used in Lancet letter |
| Date | Fri, 14 Sep 2012 16:37:19 +0000 |
This spring I added a feature to xtabond2, which offers a new way to reduce the instrument count and (one hopes) avoid the evils of instrument proliferation. The feature applies principal components analysis (PCA) to the "GMM-style" instruments and keeps only the largest components. See http://www.stata.com/statalist/archive/2012-04/msg00681.html. Today, I have a letter in the Lancet that illustrates the use of the feature. One thing I like about PCA is the way it facilitates graphical examination of the impact of reducing the instrument count on statistics such as the Hansen test. Lancet letter: http://www.thelancet.com/journals/lancet/article/PIIS0140-6736(12)61529-3/fulltext Pretty graph: http://www.thelancet.com/journals/lancet/article/PIIS0140673612615293/images?imageId=gr1§ionType=lightBlue&hasDownloadImagesLink=true Data and code: http://blogs.cgdev.org/open_book/2012/09/cross-post-the-aid-fungibility-debate-and-medical-journal-peer-review.php Articles testing PCA: Jens Mehrhoff, http://www.bundesbank.de/Redaktion/EN/Downloads/Publications/Discussion_Paper_1/2009/2009_11_06_dkp_31.pdf?__blob=publicationFile Maria Elena Bontempi and Irene Mammi, http://mpra.ub.uni-muenchen.de/40720/ --David * * 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/