Simplified Output for Matchit package

November 2, 2010

(This article by Philip Parker was originally published at Adventures in R.)

Have been attending a seminar on propensity score matching this week, when I was asked if I could write a function to reduce the complexity of the matchit package output. I produced two functions.

First I wrote a function to produce a histogram of the matched mean differences which gives a nice summary of how well the matching procedure has achieved balance (note you need to install the package ggplot2 to use this function):

meandiffplot <- function (x) {
	mdiff<-data.frame(x$sum.matched)
	names(mdiff)[4]<-c("m_mean_diff")
	diffplot<-ggplot(mdiff, aes(m_mean_diff) )
	diffplot<- diffplot+ geom_histogram (fill="grey")
	diffplot<- diffplot+ geom_density (colour="red")
	diffplot
	}

Next I produced a function the reports only those matched mean differences with a standardized difference of over .1 (sorted by absolute size).

meandifftable<- function (x){
	post<-data.frame(x$sum.matched[4])
	matchID <- as.vector (row.names (post) )
	names(post)[1]<-c("m_mean_diff")
	post$absolute<- abs(post[1])
	total2<-post[order (-post$absolute, na.last=NA) ,]
	meandiffover1<- subset(total2[1], total2[1]> .1 | total2[1]< -.1)
	meandiffover1
}