#=============================================
#code for Chapter 5 "Propensity Score Matching" of book:
#Leite, W. L. (2017). Practical propensity score methods using R. 
#Thousand Oaks, CA: Sage.  

#PART 5 - GENETIC MATCHING WITH PROPENSITY SCORES AND SENSITIVITY ANALYSIS

# 
#this is the code that was used to generate the example results in the book
#As the R software and the R packages used in this example are updated frequently
#some incompatibilities between the current code and new R versions or package versions
#may appear
#Any updates to the code will be posted at:
# http://www.practicalpropensityscore.com


#This example estimates the effect of mothers having a job 
#that provides or subsidizes child care 
#on the length that they breastfeed their children 
#National Longitudinal Survey of Youth 1979 (NLSY79) 
#and the NLSY79 Children and Youth

#perform genetic matching for estimating the ATT 
#with variable ratio matching with replacement 
#based on the propensity score only and without a caliper

#load data
load(file="Chapter5_data_with_propensity_scores_and_formula.rData")


 #library for propensity score matching
library(MatchIt)

#convert the treatment from 0/1 to TRUE/FALSE as required by GenMatch
data$childCare <- ifelse(data$childCare==1, TRUE, FALSE)

#perform matching
geneticMatching2 <- matchit(psFormula,distance=data$logitPScores, 
                        data = data, method = "genetic",pop.size=1000,
                        fit.func="pvals", 
                        estimand="ATT", replace=T, ties=T)


#diagnose covariate balance
(balance.geneticMatching2 <- summary(geneticMatching2, standardize=T))
#obtain the summary of balance aftdr matching 
summary(abs(balance.geneticMatching2$sum.matched$"Std. Mean Diff."))
table(abs(balance.geneticMatching2$sum.matched$"Std. Mean Diff.") > 0.1)

#================================================
#estimate ATT 
#using regression

#obtain matched data
data.geneticMatching2 <- match.data(geneticMatching2)

#set up the survey design
library(survey)
design.geneticMatching2 <- svydesign(ids=~1, weights=~weights,
                                     data=data.geneticMatching2)

#fit regression model to estimate the ATT
model.geneticMatching2 <- svyglm(C0338600~childCare, design.geneticMatching2, family=gaussian())
summary(model.geneticMatching2)