Practical Propensity Score Methods Using R by Walter L. Leite

Author:Walter L. Leite [Leite, Walter L.]
Language: eng
Format: epub
Publisher: SAGE Publications
Published: 2016-10-27T21:00:00+00:00

geneticMatchingBA <- Match(Y=data$C0338600, Tr=data$childCare, X=covariateData,

BiasAdjust=T, Z=covariateData, Weight.matrix = geneticWeights,

estimand = “ATT”, M = 1, replace=TRUE, ties=TRUE)

The code for multivariate genetic matching with bias adjustment for all covariates shown above produces estimates that can be extracted with the summary function:

summary(geneticMatchingBA)

The estimate of the ATT is 4.352 (SE = 2.769, p = .116), showing that bias adjustment did not alter the conclusion obtained without bias adjustment that there is no treatment effect.

The MatchIt package (Ho et al., 2011) does not estimate treatment effects directly, but it provides a matched data set with case weights that can be used to estimate the ATT. In one-to-one matching without replacement, all case weights are 1. However, if fixed ratio (greater than one-to-one) or variable ratio matching is done, weights for treated cases are 1 but weights for all untreated cases matched to a treated case are the inverse of the total number of matches the treated case received. Also, if matching is done with replacement, case weights for each untreated case are summed across the multiple matched groups in which it was included. Finally, the weights of the matched cases are multiplied by the ratio of the total number of matched cases and total number of treated cases, which scales the untreated weights to sum to the number of matched cases (Ho, Imai, King, & Stuart, 2014). The following equation implements the calculation of weights described above:

(5.7)

w i = { 1 if Z i = 1 n 0 n 1 Σ m = 1 n i 1 M m if Z i = 0

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.