Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Ariel Linden, DrPH" <ariel.linden@gmail.com> |
To | <statalist@hsphsun2.harvard.edu> |
Subject | re: st: Propensity Score Matching fore time-varying treatments |
Date | Sun, 27 Jan 2013 10:57:18 -0500 |
It is highly recommended that you look through the Statalist archives before posting a query to the list. In fact, this particular question has been asked and answered several times over the past few years. Rather than rehashing old territory, I suggest you read the post in this and its thread. I give several references that you should consider reading. http://www.stata.com/statalist/archive/2012-11/msg00594.html Ariel Date: Sat, 26 Jan 2013 22:35:55 +0100 From: Luis Aguiar <stataluis@gmail.com> Subject: st: Propensity Score Matching fore time-varying treatments Dear stata users, I am working with firm level data and want to estimate the impact of participation in a given program on firm R&D. I want to use propensity score matching in order to construct a proper control group and do a simple diff-in-diff exercise. The particularity is that not all firms enter the program in the same year. For example, firms from group A may enter the program in 2000, firms from group B may enter it in 2001, and firms from group C may enter in 2002 (but each firm participates only once). I therefore want to construct my control group as follows: 1) For firms from group A: choose non-participating firms which, in 1999 (one year before firms from group A enters the program), are as similar as possible as firms from group A in that same year. 2) For firms from group B: choose non-participating firms which, in 2000 (one year before firms from group B enters the program), are as similar as possible as firms from group B in that same year. 3) For firms from group C: choose non-participating firms which, in 2001 (one year before firms from group C enters the program), are as similar as possible as firms from group C in that same year. What I did so far was to use the psmatch2 command by Leuven and Sianesi (2003), and do the PSM separately for each year (i.e. running PSMATCH2 separately for each of the steps 1), 2) and 3) above.) I couldn't find a way of doing the matching simultaneously for all years, which would be much more convenient in order to test the overall quality of the matching (balancing tests). Does anyone know if this is possible with PSMATCH2 or with any other command? My approach may also be wrong, so any help would be greatly appreciated! I hope that my explanation is not too confusing! Best and thanks in advance, Luis * * 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/