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: Re: st: Balancing Property not satisfied |
Date | Fri, 16 Aug 2013 09:15:02 -0400 |
Vikram, Have you created a table of baseline characteristics between your two groups to see if there is overlap on those observed characteristics? Have you performed some visual checks of their distributions, such as QQ-plots, box-plots, density plots, etc.? If your covariates don't balance, it is likely that you don't have sufficient overlap in the distributions, thus, any analysis you'd perform would be resorting to extrapolation. Take for example two groups with no overlap in age. No amount of matching will get these groups to look comparable. If age is an independent predictor of the outcome, you'll have problems making the case that you've controlled for age. Since you did not share with us any tables or output, that is about as much of a response as I can offer you... There is no substitute for conducting basic analysis of the data to inform the researcher where to go next. Ariel Date: Thu, 15 Aug 2013 21:53:43 +0100 From: V!cky Finavker <vikramfinavker@gmail.com> Subject: Re: st: Balancing Property not satisfied Hi Lukas, Thanks for you help. However, when i change those covariates Pscore shows another covariates as having balancing problem. so do i have to keep on trying until i find one which satisfies or there is a solution? ______ Regards, Vikram Finavker On 15 Aug 2013, at 12:03, Lukas Borkowski <570722@soas.ac.uk> wrote: > Vikram, > > in the output window of -pscore- (a user-written command, written by Becker and Ichino, 2002) you are provided with information on which covariates are not balanced. It is written just above the dialogue you have quoted. > > Best, > > Lukas > Becker, S. and A. Ichino (2002): Estimation of average treatment effects based on propensity scores, The Stata Journal 2(4): pp. 358-377. > > # * * 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/