Gary King Homepage Previous: The plot() Command Up: User's Guide to MATCHIT Next: Quick Overview


Conducting Analyses after Matching

Any software package may be used for parametric analysis following MATCHIT. This includes any of the relevant R packages, or other statistical software by exporting the resulting matched data sets using R commands such as write.csv() and write.table() for ASCII files or write.dta() in the foreign package for a STATA binary file.

When variable numbers of treated and control units have been matched to each other (e.g., through exact matching, full matching, or k:1 matching with replacement), the weights created by MatchIt should be used (e.g., in a weighted regression) to ensure that the matched treated and control groups are weighted up to be similar. Users should also remember that the weights created by MatchIt estimate the average treatment effect on the treated, with the control units weighted to resemble the treated units. See below for more detail on the weights. With subclassification, estimates should be obtained within each subclass and then aggregated across subclasses. When it is not possible to calculate an effect within each subclass, again the weights can be used to weight the matched units.

In this section, we show how to use Zelig with MATCHIT. Zelig (Imai et al., 2006) is an R package that implements a large variety of statistical models (using numerous existing R packages) with a single easy-to-use interface, gives easily interpretable results by simulating quantities of interest, provides numerical and graphical summaries, and is easily extensible to include new methods.



Subsections

Gary King 2010-12-11