Matching with Clustered Data: the CMatching Package in R

Authors

Massimo Cannas, University of Cagliari
Bruno Arpino, University of Florence

Date of this Version

6-2019

Document Type

Article

Citation

The R Journal (June 2019) 11(1); Editor: Norm Matloff

Comments

Copyright 2019, The R Foundation. Open access material. License: CC BY 4.0 International

Abstract

Matching is a well known technique to balance covariates distribution between treated and control units in non-experimental studies. In many fields, clustered data are a very common occurrence in the analysis of observational data and the clustering can add potentially interesting information. Matching algorithms should be adapted to properly exploit the hierarchical structure. In this article we present the CMatching package implementing matching algorithms for clustered data. The package provides functions for obtaining a matched dataset along with estimates of most common parameters of interest and model-based standard errors. A propensity score matching analysis, relating math proficiency with homework completion for students belonging to different schools (based on the NELS-88 data), illustrates in detail the use of the algorithms.