Any help on the following would be greatly appreciated.
I have a panel with treated and controls (potential twins) in which
treatment occurs in different years and twins should be found in the
same year as treatment, so I cannot simply pool the data.
My first idea was to estimate propensity scores for each year separately
(using 'pscore'), then adding the propensity scores together and
thereafter estimating the att (using 'attnd') while taking the bootstrap
sample from respective year using strata(year) as:
attnd outcome treat, pscore(mypscore) comsup boot strata(year).
An additional problem is that in some years only one individual is
treated leading to a singleton treatment dummy.
Any comments on why this might be considered crazy and how to proceed?
Best wishes,
Jan Pettersson
*
* 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/
