| Previous: Introduction | Up: Introduction | Next: Software Requirements |
MATCHIT implements the suggestions of Ho, Imai, King, and Stuart (2007) for improving parametric statistical models and reducing model dependence by preprocessing data with semi-parametric and non-parametric matching methods. After appropriately preprocessing with MATCHIT, researchers can use whatever parametric model and software they would have used without MATCHIT, without other modification, and produce inferences that are more robust and less sensitive to modeling assumptions. (In addition, you may wish to use Zelig (http://gking.harvard.edu/zelig/; Imai et al. 2006 for subsequent parametric analyses, as it is designed to be convenient in analyzing MATCHIT data sets.) MATCHIT reduces the dependence of causal inferences on commonly made, but hard-to-justify, statistical modeling assumptions via the largest range of sophisticated matching methods of any software we know of. The program includes most existing approaches to matching and even enables users to access methods implemented in other programs through its single, unified, and easy-to-use interface. In addition, we have written MATCHIT so that adding new matching methods to the software is as easy for anyone with the inclination as it is for us.