re: st: propensity score matching with panel data

["Ariel Linden. DrPH" <ariel.linden@gmail.com>]

"matching" in the traditional meaning of the word does not make much sense
here, since treatment is time-varying. A more appropriate approach would be
to use "weighting" where each individual's probability of being treated at
time T can be estimated. See the reference below, which do exactly that...

Hernán, M. A., Brumback, B. & Robins, J. M. (2002) Estimating the causal
effect of zidovudine on CD4 count with a marginal structural model for
repeated measures. Statistics in Medicine, 21, 1689?1709.

Ariel
________________________________________
From
  "superabu@libero.it" <superabu@libero.it>
To
  statalist@hsphsun2.harvard.edu
Subject
  st: propensity score matching with panel data
Date
  Thu, 11 Jul 2013 18:49:35 +0200 (CEST)
________________________________________
Hi,

 

I have a question on propensity score matching methodology.


I have a panel of 73 countires and 5 periods (5 years average of originally
25 
years).
I am using propensity score matching to test the impact of a specific
program 
on fiscal behaviour.

The treatment is : being under the program.
(i.e. a country can be under a program in the first period, not beign under
a 
program in the second period and then being again under the program in the 
third period)
 
The outcome variable is a proxy of fiscal behavior.

I am using a panel probit to estimate the propensity score. 

Now, given the importance of the time dimension in my analysis I was
wondering 
whether it makes sense to include time dummy variable in this stage? 
or I should matched (in the second stage) taking into account periods?

 

Thanks a lot for you help,



*
*   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/