Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: Propensity Score Matching (PSM) - matching problem

From   umut senalp <>
To   <>
Subject   RE: st: Propensity Score Matching (PSM) - matching problem
Date   Thu, 6 Sep 2012 19:06:53 +0300

Dear Ariel, 

Thank you for your reply. The treatment status of the firms may change over the time as you suggested, but I guess I will be able to solve my problem by restricting a firm to be considered as "treated" only 1 time during the period it was observed, and I will code the treatment dummy as missing variable for the period following the treatment. By doing this I hope I'll manage to prevent the unwanted matches that I tried to describe in my previous post. Thank you again for your suggestions.



> From:
> To:
> Subject: re: st: Propensity Score Matching (PSM) - matching problem
> Date: Thu, 6 Sep 2012 05:41:33 -0400
> The only reason I can think of that you'd have this issue arise is if your
> treatment group is not consistently coded as 1 and your non-treatment group
> is not consistently coded as 0.
> There are basically two ways to handle "matched" data over time. The first
> would entail matching (or weighting) on pre-intervention data and assume
> that the treatment status never changes over time. In this case, you'd
> control for pre-intervention covariates by matching (or weighting) and not
> do any further adjustments after the intervention starts.
> The second would assume that the treatment status could possibly change over
> time (as well as allowing for time-varying covariates). In health-care this
> could happen if a treatment (let's say a drug) is given at some point, and
> then not given, and then given again. This is obviously a much more
> complicated model to deal with than the first scenario (See Hernán et al*
> for a good description of such an approach)...
> I am not sure this is the issue with your data, but it seems to be a
> reasonable assumption based on the limited information you provided....
> I hope this helps
> Ariel
> * 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. 
> Date: Thu, 6 Sep 2012 02:40:27 +0300
> From: umut senalp <>
> Subject: st: Propensity Score Matching (PSM) - matching problem
> Dear Statalisters,
> I am currently working with a panel containing around 9.000 firms and
> trying to evaluate the exports effects on firm-level productivity, and
> I use Matching approach (Propensity Score Matching) to get the
> treatment/export effects (on treated - ATT).
> I use -psmatch2.ado- (writen by Edwin Leuven and Barbara Sianesi) 
> I already matched the exporters with non exporters. However, I am
> having some problems with the application of the method. The problem 
> is PSM procedure allows the PSM algorithm to match
> a treated firm with a treatment-group firm (and sometime itself) 
> after the treatment is over, by considering
> treated firmsafter treatment as if they were control firms.
> What I would like to ask is if is there any solution about the problem.
> I would be glad if you could provide me a hint on this issue.
> Kind regards 
> *
> * For searches and help try:
> *
> *
> *
*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index