There different because they match treated obs to control obs in different ways and can, therefore, generate very different answers. There is no basis for averaging the various estimates. Nearest neighbor matching minimizes the bias since the only the nearest control is used, while kernel matching is more efficient since it uses all controls (and I believe better in a mean-squared error sense). See the recent Journal of Econometrics article by Petra Todd and Jeff Smith. Also, there is a recent IZA discussion paper on "Practical Guidelines to Matching" or something close to that. Don't recall the author.
Dann
****************************************************
Daniel L. Millimet, Associate Professor
Department of Economics
Box 0496
SMU
Dallas, TX USA
phone: 214.768.3269
fax: 214.768.1821
web: http://faculty.smu.edu/millimet
****************************************************
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Manuel Ch�vez
Sent: Monday, July 18, 2005 10:20 AM
To: [email protected]
Subject: st: psmatch and matching methods
I'm using psmatch to estimate the matching of several variables and I'm
obtaining very different result on the impact depending if I use the nearest
neighbor, the Mahlanabis distance and the kernel matching methods. My
questions are:
1) If somebody knows why this happen?
2) If it is correct to take the average of the three methods for obtaining a
more acurate estimation of the impacts?
Thanks
Manuel G. Ch�vez Angeles
FAO
Consultor Nacional de Impactos
Providencia No. 334 5� Piso
Col. del Valle
M�xico, D.F. 03100
M�xico
Tel. 11076425/26/30/31 Ext. 319
*
* 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/
*
* 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/
