Overview
MatchIt provides a simple and straightforward interface to various methods of matching for covariate balance in observational studies. Matching is one way to reduce confounding and model dependence when estimating treatment effects. Several matching methods are available, including nearest neighbor matching, optimal pair matching, optimal full matching, genetic matching, exact matching, coarsened exact matching, and subclassification, some of which rely on functions from other R packages. A variety of methods to estimate propensity scores for propensity score matching are included. Below is an example of the use of MatchIt to perform 2:1 nearest neighbor propensity score matching with a propensity score caliper and assessing overlap and balance:
library("MatchIt")
data("lalonde", package = "MatchIt")
#Nearest neighbor PS matching without replacement and with a caliper
m.out <- matchit(treat ~ age + educ + race + married + nodegree + re74 + re75,
data = lalonde, ratio = 2, caliper = .025)Printing the MatchIt object provides details of the kind of matching performed.
m.out#> A matchit object
#> - method: 2:1 nearest neighbor matching without replacement
#> - distance: Propensity score [caliper]
#> - estimated with logistic regression
#> - caliper: <distance> (0.007)
#> - number of obs.: 614 (original), 234 (matched)
#> - target estimand: ATT
#> - covariates: age, educ, race, married, nodegree, re74, re75We can view propensity score overlap and see which observations were matched and unmatched using a jitter plot:
#Checking for PS overlap
plot(m.out, type = "jitter", interactive = FALSE)
With this we can see that most of the unmatched control units had small propensity scores, making them unlike the treated group. Other plots are available to view the distributions of propensity scores and covariates.
We can check covariate balance for the original and matched samples using summary():
#Checking balance before and after matching:
summary(m.out)#>
#> Call:
#> matchit(formula = treat ~ age + educ + race + married + nodegree +
#> re74 + re75, data = lalonde, caliper = 0.025, ratio = 2)
#>
#> Summary of Balance for All Data:
#> Means Treated Means Control Std. Mean Diff. Var. Ratio eCDF Mean eCDF Max
#> distance 0.5774 0.1822 1.7941 0.9211 0.3774 0.6444
#> age 25.8162 28.0303 -0.3094 0.4400 0.0813 0.1577
#> educ 10.3459 10.2354 0.0550 0.4959 0.0347 0.1114
#> raceblack 0.8432 0.2028 1.7615 . 0.6404 0.6404
#> racehispan 0.0595 0.1422 -0.3498 . 0.0827 0.0827
#> racewhite 0.0973 0.6550 -1.8819 . 0.5577 0.5577
#> married 0.1892 0.5128 -0.8263 . 0.3236 0.3236
#> nodegree 0.7081 0.5967 0.2450 . 0.1114 0.1114
#> re74 2095.5737 5619.2365 -0.7211 0.5181 0.2248 0.4470
#> re75 1532.0553 2466.4844 -0.2903 0.9563 0.1342 0.2876
#>
#>
#> Summary of Balance for Matched Data:
#> Means Treated Means Control Std. Mean Diff. Var. Ratio eCDF Mean eCDF Max Std. Pair Dist.
#> distance 0.4962 0.4937 0.0115 1.0104 0.0043 0.0392 0.0131
#> age 25.3627 25.5637 -0.0281 0.4399 0.0800 0.2157 1.3150
#> educ 10.2549 10.4216 -0.0829 0.6076 0.0237 0.0784 1.0060
#> raceblack 0.7157 0.7157 0.0000 . 0.0000 0.0000 0.0379
#> racehispan 0.1078 0.1078 0.0000 . 0.0000 0.0000 0.1439
#> racewhite 0.1765 0.1765 0.0000 . 0.0000 0.0000 0.1364
#> married 0.2059 0.2157 -0.0250 . 0.0098 0.0098 0.6190
#> nodegree 0.6765 0.6422 0.0755 . 0.0343 0.0343 0.7832
#> re74 2409.1384 2573.6987 -0.0337 1.4815 0.0528 0.2598 0.7939
#> re75 1704.2015 1721.3424 -0.0053 1.6335 0.0391 0.1275 0.8505
#>
#> Percent Balance Improvement:
#> Std. Mean Diff. Var. Ratio eCDF Mean eCDF Max
#> distance 99.4 87.3 98.9 93.9
#> age 90.9 -0.0 1.6 -36.7
#> educ -50.8 29.0 31.6 29.6
#> raceblack 100.0 . 100.0 100.0
#> racehispan 100.0 . 100.0 100.0
#> racewhite 100.0 . 100.0 100.0
#> married 97.0 . 97.0 97.0
#> nodegree 69.2 . 69.2 69.2
#> re74 95.3 40.2 76.5 41.9
#> re75 98.2 -998.1 70.8 55.7
#>
#> Sample Sizes:
#> Control Treated
#> All 429. 185
#> Matched (ESS) 119.59 102
#> Matched 132. 102
#> Unmatched 297. 83
#> Discarded 0. 0At the top is balance for the original sample. Below that is balance in the matched sample, followed by the percent reduction in imbalance and the sample sizes before and after matching. Smaller values for the balance statistics indicate better balance. We can plot the standardized mean differences in a Love plot for a clean, visual display of balance across the sample:
Although much has been written about matching theory, most of the theory relied upon in MatchIt is described well in Ho, Imai, King, and Stuart (2007) and Stuart (2010). The Journal of Statistical Software article for MatchIt can be accessed here, though note that some options have changed, so the MatchIt reference pages and included vignettes should be used for understanding the functions and methods available. Further references for individual methods are present in their respective help pages. The MatchIt website provides access to vignettes and documentation files.
Citing MatchIt
Please cite MatchIt when using it for analysis presented in publications, which you can do by citing the Journal of Statistical Software article below:
Ho, D. E., Imai, K., King, G., & Stuart, E. A. (2011). MatchIt: Nonparametric Preprocessing for Parametric Causal Inference. Journal of Statistical Software, 42(8). doi:10.18637/jss.v042.i08
This citation can also be accessed using citation("MatchIt") in R. For reproducibility purposes, it is also important to include the version number for the version used.