In ngreifer/lmw: Linear Model Weights
knitr::opts_chunk$set(echo = TRUE)
Here we benchmark lmw to results from other procedures and packages to ensure lmw is working as expected.
library(lmw)
data("lalonde")
lalonde$treatf <- factor(lalonde$treat) #factor version of treat for marginaleffects()
lalonde$sw <- runif(nrow(lalonde)) #sampling weights
library(lmtest); library(sandwich);
library(marginaleffects); library(estimatr)
library(fixest)
No Weights
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATE
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
estimand = "ATE")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### marginaleffects() output
f <- lm(re78 ~ treatf * (age + educ + race + married + re74 + re75),
data = lalonde)
summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f)))
### estimatr output
fl <- estimatr::lm_lin(re78 ~ treat,
~ age + educ + race + married + re74 + re75,
data = lalonde, se_type = "HC3")
summary(fl)$coefficients["treat",,drop=FALSE]
MRI, binary treatment, ATT
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
estimand = "ATT")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### marginaleffects() output
f <- lm(re78 ~ treatf * (age + educ + race + married + re74 + re75),
data = lalonde)
summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f),
newdata = subset(lalonde, treat == 1)))
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat",
fixef = ~race)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race,
data = lalonde)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
### lm_robust() output
fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, fixed_effects = ~race,
se_type = "HC3")
coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATE, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
fixef = ~race)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### URI + covariate version
l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race,
data = lalonde, method = "URI", treat = "treat")
weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0])
summary(lmw_est(l2))
### lm() output
f <- lm(re78 ~ treatf * (age + educ + married + re74 + re75) + race,
data = lalonde)
summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f),
data = lalonde))
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, treat = "treat",
iv = ~nodegree + race)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1"))
### feols() output
f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race,
data = lalonde, vcov = "HC1")
coeftest(f)["fit_treat",,drop=F]
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 |
nodegree + race + age + educ + married + re74 + re75,
data = lalonde, se_type = "HC1")
coeftest(fl)["treat",,drop=F]
URI, binary treatment, 2SLS, fixed effects, CR SEs
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, treat = "treat",
iv = ~nodegree, fixef = ~race)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1", cluster = ~race))
### feols() output
f <- feols(re78 ~ age + educ + married + re74 + re75 + race | treat ~ nodegree,
data = lalonde, cluster = ~race)
coeftest(f)["fit_treat",,drop=F]
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 |
nodegree + race + age + educ + married + re74 + re75,
data = lalonde, fixed_effects = ~race,
se_type = "stata", cluster = race)
coeftest(fl)["treat",,drop=F]
#Note: SEs agree for cluster HC1 (aka iv_robust() "stata" and feols() "cluster") and
# HC2 (aka iv_robust() "CR2") but not HC0 (aka iv_robust() "CR0"). feols() only
# has HC1. Full agreement with vcovCL() used on ivreg::ivreg().
MRI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ,
data = lalonde, treat = "treat", method = "MRI",
iv = ~nodegree + race)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1"))
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat * (scale(age, s=F) + scale(educ, s=F)) |
(nodegree + race) * (scale(age, s=F) + scale(educ, s=F)),
data = lalonde, se_type = "HC1")
coeftest(fl)["treat",,drop=F]
URI Regression, multi-category treatment
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "race",
contrast = c("hispan", "white"))
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race == "hispan"]) -
weighted.mean(lalonde$re78[lalonde$race != "hispan"], l$weights[lalonde$race != "hispan"])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ relevel(race, "white") + age + educ + married + re74 + re75,
data = lalonde)
coeftest(f, vcov. = vcovHC)[3,,drop=FALSE]
MRI Regression, multi-category treatment, ATE
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "race",
estimand = "ATE")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"]) -
weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"])
### lmw_est() output
summary(lmw_est(l))$coef
### marginaleffects() output
f <- lm(re78 ~ race * (age + educ + married + re74 + re75),
data = lalonde)
summary(marginaleffects(f, var = "race", vcov = vcovHC(f)))
### estimatr output
fl <- estimatr::lm_lin(re78 ~ race,
~ age + educ + married + re74 + re75,
data = lalonde, se_type = "HC3")
summary(fl)$coefficients[1:3,]
MRI Regression, multi-category treatment, ATT
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "race",
estimand = "ATT", focal = "black")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race=="hispan"]) -
weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"])
### lmw_est() output
summary(lmw_est(l))$coef
### marginaleffects() output
f <- lm(re78 ~ race * (age + educ + married + re74 + re75),
data = lalonde)
summary(marginaleffects(f, var = "race", vcov = vcovHC(f),
newdata = subset(lalonde, race == "black")))
URI Regression, multi-category treatment, fixed effects
l <- lmw(re78 ~ race + age + educ + re74 + re75,
data = lalonde, method = "URI", treat = "race",
contrast = c("hispan", "white"), fixef = ~married)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race == "hispan"]) -
weighted.mean(lalonde$re78[lalonde$race != "hispan"], l$weights[lalonde$race != "hispan"])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ relevel(race, "white") + age + educ + married + re74 + re75,
data = lalonde)
coeftest(f, vcov. = vcovHC)[3,,drop=FALSE]
### estimatr output
fl <- estimatr::lm_robust(re78 ~ relevel(race, "white") + age + educ + re74 + re75,
data = lalonde, fixed_effects = ~married,
se_type = "HC3")
coeftest(fl)[2,,drop = FALSE]
Sampling Weights
lalonde$sw <- runif(nrow(lalonde))
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat",
s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, weights = sw)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATE
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
estimand = "ATE", s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### estimatr output
fl <- estimatr::lm_lin(re78 ~ treat,
~ age + educ + race + married + re74 + re75,
data = lalonde, se_type = "HC3",
weights = sw)
summary(fl)$coefficients["treat",,drop=FALSE]
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat",
fixef = ~race, s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race,
data = lalonde, weights = sw)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
### lm_robust() output
fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, fixed_effects = ~race,
se_type = "HC3", weights = sw)
coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATE, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
fixef = ~race, s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### URI + covariate version
l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race,
data = lalonde, method = "URI", treat = "treat", s.weights = sw)
weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0])
summary(lmw_est(l2))$coef
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, treat = "treat",
iv = ~nodegree + race, s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1"))$coef
### feols() output
f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race,
data = lalonde, vcov = "HC1", weights = ~sw)
coeftest(f)["fit_treat",,drop=F]
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 |
nodegree + race + age + educ + married + re74 + re75,
data = lalonde, se_type = "HC1", weights = sw)
coeftest(fl)["treat",,drop=F]
URI, binary treatment, 2SLS, fixed effects, CR SEs
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, treat = "treat",
iv = ~nodegree, fixef = ~race, s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1", cluster = ~race))$coef
### feols() output
f <- feols(re78 ~ age + educ + married + re74 + re75 + race | treat ~ nodegree,
data = lalonde, cluster = ~race, weights= ~sw)
coeftest(f)["fit_treat",,drop=F]
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 |
nodegree + race + age + educ + married + re74 + re75,
data = lalonde, fixed_effects = ~race,
se_type = "stata", cluster = race, weights = sw)
coeftest(fl)["treat",,drop=F]
#Note: SEs agree for cluster HC1 (aka iv_robust() "stata" and feols() "cluster") and
# HC2 (aka iv_robust() "CR2") but not HC0 (aka iv_robust() "CR0"). feols() only
# has HC1. Full agreement with vcovCL() used on ivreg::ivreg().
MRI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ,
data = lalonde, treat = "treat", method = "MRI",
iv = ~nodegree, s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1"))$coef
### iv_robust()
age_c_sw <- lalonde$age - weighted.mean(lalonde$age, lalonde$sw)
educ_c_sw <- lalonde$educ - weighted.mean(lalonde$educ, lalonde$sw)
fl <- estimatr::iv_robust(re78 ~ treat * (age_c_sw + educ_c_sw) |
nodegree * (age_c_sw + educ_c_sw),
data = lalonde, se_type = "HC1", weights = sw)
coeftest(fl)["treat",,drop=F]
URI Regression, multi-category treatment
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "race",
contrast = c("hispan", "white"), s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race == "hispan"]) -
weighted.mean(lalonde$re78[lalonde$race != "hispan"], l$weights[lalonde$race != "hispan"])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ relevel(race, "white") + age + educ + married + re74 + re75,
data = lalonde, weights = sw)
coeftest(f, vcov. = vcovHC)[3,,drop=FALSE]
MRI Regression, multi-category treatment, ATE
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "race",
estimand = "ATE", s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) -
weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"])
### lmw_est() output
summary(lmw_est(l))$coef
### estimatr output
fl <- estimatr::lm_lin(re78 ~ race,
~ age + educ + married + re74 + re75,
data = lalonde, se_type = "HC3", weights = sw)
summary(fl)$coefficients[1:3,]
URI Regression, multi-category treatment, fixed effects
l <- lmw(re78 ~ race + age + educ + re74 + re75,
data = lalonde, method = "URI", treat = "race",
contrast = c("hispan", "white"), fixef = ~married,
s.weights = sw)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race == "hispan"]) -
weighted.mean(lalonde$re78[lalonde$race != "hispan"], l$weights[lalonde$race != "hispan"])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ relevel(race, "white") + age + educ + married + re74 + re75,
data = lalonde, weights = sw)
coeftest(f, vcov. = vcovHC)[3,,drop=FALSE]
### estimatr output
fl <- estimatr::lm_robust(re78 ~ relevel(race, "white") + age + educ + re74 + re75,
data = lalonde, fixed_effects = ~married,
se_type = "HC3", weights = sw)
coeftest(fl)[2,,drop = FALSE]
Base Weights from MatchIt
library(MatchIt)
M <- matchit(treat ~ age + educ + race + married + re74 + re75, data = lalonde,
method = "nearest", estimand = "ATT")
md <- match.data(M, data = lalonde)
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat",
base.weights = M$weights)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75,
data = md, weights = weights)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATT
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
estimand = "ATT", base.weights = M$weights)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### marginaleffects() output
f <- lm(re78 ~ treatf * (age + educ + race + married + re74 + re75),
data = md, weights = weights)
summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f),
newdata = subset(md, treat == 1)))
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat",
fixef = ~race, base.weights = M$weights)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race,
data = md, weights = weights)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
### lm_robust() output
fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75,
data = md, fixed_effects = ~race,
se_type = "HC3", weights = weights)
coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATT, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
fixef = ~race, estimand = "ATT", base.weights = M$weights)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### URI + covariate version
l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race,
data = lalonde, method = "URI", treat = "treat",
estimand = "ATT", base.weights = M$weights)
weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0])
summary(lmw_est(l2))$coef
### lm() output
f <- lm(re78 ~ treatf * (age + educ + married + re74 + re75) + race,
data = md, weights = weights)
summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f),
newdata = subset(md, treat == 1)))
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, treat = "treat",
iv = ~nodegree + race, base.weights = M$weights)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1"))$coef
### feols() output
f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race,
data = md, vcov = "HC1", weights = ~weights)
coeftest(f)["fit_treat",,drop=F]
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 |
nodegree + race + age + educ + married + re74 + re75,
data = md, se_type = "HC1", weights = weights)
coeftest(fl)["treat",,drop=F]
URI, binary treatment, 2SLS, fixed effects, CR SEs
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, treat = "treat",
iv = ~nodegree, fixef = ~race, base.weights = M$weights)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1", cluster = ~race))$coef
### feols() output
f <- feols(re78 ~ age + educ + married + re74 + re75 + race | treat ~ nodegree,
data = md, cluster = ~race, weights = ~weights)
coeftest(f)["fit_treat",,drop=F]
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 |
nodegree + race + age + educ + married + re74 + re75,
data = md, fixed_effects = ~race,
se_type = "stata", cluster = race, weights = weights)
coeftest(fl)["treat",,drop=F]
#Note: SEs agree for cluster HC1 (aka iv_robust() "stata" and feols() "cluster") and
# HC2 (aka iv_robust() "CR2") but not HC0 (aka iv_robust() "CR0"). feols() only
# has HC1. Full agreement with vcovCL() used on ivreg::ivreg().
Base Weights from WeightIt
library(WeightIt)
#Binary treatment, ATE
W_ate <- weightit(treat ~ age + educ + race + married + re74 + re75, data = lalonde,
method = "ps", estimand = "ATE")
#Binary treatment, ATT
W_att <- weightit(treat ~ age + educ + race + married + re74 + re75, data = lalonde,
method = "ps", estimand = "ATT")
#Multi-category treatment, ATE
#install.packages("mclogit")
W3 <- weightit(race ~ age + educ + married + re74 + re75, data = lalonde,
method = "ps", estimand = "ATE", use.mclogit = TRUE,
include.obj = TRUE)
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat", obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, weights = W_ate$weights)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATE
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat", obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### marginaleffects() output
f <- lm(re78 ~ treatf * (age + educ + race + married + re74 + re75),
data = lalonde, weights = W_ate$weights)
summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f)))
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat",
fixef = ~race, obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race,
data = lalonde, weights = W_ate$weights)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
### lm_robust() output
fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, fixed_effects = ~race,
se_type = "HC3", weights = W_ate$weights)
coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATE, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
fixef = ~race, obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### URI + covariate version
l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race,
data = lalonde, method = "URI", treat = "treat", obj = W_ate)
weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0])
summary(lmw_est(l2))$coef
### lm() output
f <- lm(re78 ~ treatf * (age + educ + married + re74 + re75) + race,
data = lalonde, weights = W_ate$weights)
summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f)))
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, treat = "treat",
iv = ~nodegree + race, obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1"))$coef
### feols() output
f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race,
data = lalonde, vcov = "HC1", weights = W_ate$weights)
coeftest(f)["fit_treat",,drop=F]
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 |
nodegree + race + age + educ + married + re74 + re75,
data = lalonde, se_type = "HC1", weights = W_ate$weights)
coeftest(fl)["treat",,drop=F]
MRI, binary treatment, ATE, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat", obj = W_ate,
dr.method = "AIPW")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### PSweight() output
P <- PSweight::PSweight(ps.estimate = W_ate$ps, zname = "treat", yname = "re78",
data = lalonde, weight = "IPW", augmentation = TRUE,
out.formula = re78 ~ age + educ + race + married + re74 + re75)
summary(P)
MRI, binary treatment, ATT, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat", obj = W_att,
dr.method = "AIPW")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### PSweight() output
P <- PSweight::PSweight(ps.estimate = W_att$ps, zname = "treat", yname = "re78",
data = lalonde, weight = "treated", augmentation = TRUE,
out.formula = re78 ~ age + educ + race + married + re74 + re75)
summary(P)
#Note: SEs don't agree for ATT; PSweight's will be a little smaller. See lmw_est.lmw_aipw
#for details. PSweight's are more trustworthy because they use a more accurate formula
#that involves the propensity score (but it requires a propensity score!). Some
#evidence that analytical SEs perform poorly with incorrect outcome model anyway;
#Mao, LI, & Greene (2018, SMMR)
URI, binary treatment, ATE, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat", obj = W_ate,
dr.method = "AIPW")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### PSweight() output
fit <- lm(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde)
ll0 <- lalonde; ll0$treat <- 0
ll1 <- lalonde; ll1$treat <- 1
y0 <- predict(fit, newdata = ll0)
y1 <- predict(fit, newdata = ll1)
P <- PSweight::PSweight(ps.estimate = W_ate$ps, zname = "treat", yname = "re78",
data = lalonde, weight = "IPW", augmentation = TRUE,
out.estimate = cbind(`0` = y0, `1` = y1))
summary(P)
#Note: PSweight SE doesn't incorporate estimation of outcome model
# because it only supports MRI natively. Point estimates are
# correct though.
MRI, multi-category treatment, ATE, AIPW
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "race", obj = W3,
dr.method = "AIPW")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) -
weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"])
### lmw_est() output
summary(lmw_est(l))$coef
### PSweight() output
ps_mat <- fitted(W3$obj); colnames(ps_mat) <- levels(lalonde$race)
P <- PSweight::PSweight(ps.estimate = ps_mat, zname = "race", yname = "re78",
data = lalonde, weight = "IPW", augmentation = TRUE,
out.formula = re78 ~ age + educ + married + re74 + re75)
summary(P)
URI, multi-category treatment, ATE, AIPW
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "race", obj = W3,
dr.method = "AIPW", contrast = c("white", "black"))
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) -
weighted.mean(lalonde$re78[lalonde$race != "white"], l$weights[lalonde$race!="white"])
### lmw_est() output
summary(lmw_est(l))$coef
### PSweight() output
ps_mat <- fitted(W3$obj); colnames(ps_mat) <- levels(lalonde$race)
fit <- lm(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde)
llb <- lalonde; llb$race[] <- "black"
llh <- lalonde; llh$race[] <- "hispan"
llw <- lalonde; llw$race[] <- "white"
yb <- predict(fit, newdata = llb)
yh <- predict(fit, newdata = llh)
yw <- predict(fit, newdata = llw)
P <- PSweight::PSweight(ps.estimate = ps_mat, zname = "race", yname = "re78",
data = lalonde, weight = "IPW", augmentation = TRUE,
out.estimate = cbind(black = yb, hispan = yh, white = yw))
summary(P)
Sampling Weights + Base Weights from WeightIt
library(WeightIt)
#Binary treatment, ATE
W_ate <- weightit(treat ~ age + educ + race + married + re74 + re75, data = lalonde,
method = "ps", estimand = "ATE", s.weights = lalonde$sw)
#Binary treatment, ATT
W_att <- weightit(treat ~ age + educ + race + married + re74 + re75, data = lalonde,
method = "ps", estimand = "ATT", s.weights = lalonde$sw)
#Multi-category treatment, ATE
#install.packages("mclogit")
W3 <- weightit(race ~ age + educ + married + re74 + re75, data = lalonde,
method = "ps", estimand = "ATE", use.mclogit = TRUE,
include.obj = TRUE, s.weights = lalonde$sw)
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat", obj = W_ate)
#Note: sampling weights and base weights are both in W_ate
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, weights = W_ate$weights * lalonde$sw)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATE
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat", obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
# Note: not straightforward to compute marginal effects using another package;
# need to take weighted average of predicted values where weights are
# only the smapling weights, but model is fit using product of sampling
# and base weights. marginaleffects() cannot do this. One option is to
# center the covariates first at sampling-weighted means.
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat",
fixef = ~race, obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### lm() output
f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race,
data = lalonde, weights = W_ate$weights * lalonde$sw)
coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
### lm_robust() output
fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, fixed_effects = ~race,
se_type = "HC3", weights = W_ate$weights * lalonde$sw)
coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATE, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat",
fixef = ~race, obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
### URI + covariate version
l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race,
data = lalonde, method = "URI", treat = "treat", obj = W_ate)
weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0])
summary(lmw_est(l2))$coef
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75,
data = lalonde, treat = "treat",
iv = ~nodegree + race, obj = W_ate)
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l, robust = "HC1"))$coef
### feols() output
f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race,
data = lalonde, vcov = "HC1", weights = W_ate$weights * lalonde$sw)
coeftest(f)["fit_treat",,drop=F]
### iv_robust()
fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 |
nodegree + race + age + educ + married + re74 + re75,
data = lalonde, se_type = "HC1",
weights = W_ate$weights * lalonde$sw)
coeftest(fl)["treat",,drop=F]
MRI, binary treatment, ATE, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat", obj = W_ate,
dr.method = "AIPW")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
MRI, binary treatment, ATT, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "MRI", treat = "treat", obj = W_att,
dr.method = "AIPW")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
URI, binary treatment, ATE, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75,
data = lalonde, method = "URI", treat = "treat", obj = W_ate,
dr.method = "AIPW")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) -
weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0])
### lmw_est() output
summary(lmw_est(l))$coef
MRI, multi-category treatment, ATE, AIPW
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "MRI", treat = "race", obj = W3,
dr.method = "AIPW")
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) -
weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"])
### lmw_est() output
summary(lmw_est(l))$coef
URI, multi-category treatment, ATE, AIPW
l <- lmw(re78 ~ race + age + educ + married + re74 + re75,
data = lalonde, method = "URI", treat = "race", obj = W3,
dr.method = "AIPW", contrast = c("white", "black"))
### Weighted difference in means
weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) -
weighted.mean(lalonde$re78[lalonde$race != "white"], l$weights[lalonde$race!="white"])
### lmw_est() output
summary(lmw_est(l))$coef
ngreifer/lmw documentation built on Feb. 14, 2024, 10:53 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
knitr::opts_chunk$set(echo = TRUE)
Here we benchmark lmw to results from other procedures and packages to ensure lmw is working as expected.
library(lmw) data("lalonde") lalonde$treatf <- factor(lalonde$treat) #factor version of treat for marginaleffects() lalonde$sw <- runif(nrow(lalonde)) #sampling weights library(lmtest); library(sandwich); library(marginaleffects); library(estimatr) library(fixest)
No Weights
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "URI", treat = "treat") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATE
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", estimand = "ATE") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### marginaleffects() output f <- lm(re78 ~ treatf * (age + educ + race + married + re74 + re75), data = lalonde) summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f))) ### estimatr output fl <- estimatr::lm_lin(re78 ~ treat, ~ age + educ + race + married + re74 + re75, data = lalonde, se_type = "HC3") summary(fl)$coefficients["treat",,drop=FALSE]
MRI, binary treatment, ATT
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", estimand = "ATT") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### marginaleffects() output f <- lm(re78 ~ treatf * (age + educ + race + married + re74 + re75), data = lalonde) summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f), newdata = subset(lalonde, treat == 1)))
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", fixef = ~race) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race, data = lalonde) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE] ### lm_robust() output fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, fixed_effects = ~race, se_type = "HC3") coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATE, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", fixef = ~race) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### URI + covariate version l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race, data = lalonde, method = "URI", treat = "treat") weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0]) summary(lmw_est(l2)) ### lm() output f <- lm(re78 ~ treatf * (age + educ + married + re74 + re75) + race, data = lalonde) summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f), data = lalonde))
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, treat = "treat", iv = ~nodegree + race) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1")) ### feols() output f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race, data = lalonde, vcov = "HC1") coeftest(f)["fit_treat",,drop=F] ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 | nodegree + race + age + educ + married + re74 + re75, data = lalonde, se_type = "HC1") coeftest(fl)["treat",,drop=F]
URI, binary treatment, 2SLS, fixed effects, CR SEs
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, treat = "treat", iv = ~nodegree, fixef = ~race) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1", cluster = ~race)) ### feols() output f <- feols(re78 ~ age + educ + married + re74 + re75 + race | treat ~ nodegree, data = lalonde, cluster = ~race) coeftest(f)["fit_treat",,drop=F] ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 | nodegree + race + age + educ + married + re74 + re75, data = lalonde, fixed_effects = ~race, se_type = "stata", cluster = race) coeftest(fl)["treat",,drop=F] #Note: SEs agree for cluster HC1 (aka iv_robust() "stata" and feols() "cluster") and # HC2 (aka iv_robust() "CR2") but not HC0 (aka iv_robust() "CR0"). feols() only # has HC1. Full agreement with vcovCL() used on ivreg::ivreg().
MRI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ, data = lalonde, treat = "treat", method = "MRI", iv = ~nodegree + race) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1")) ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat * (scale(age, s=F) + scale(educ, s=F)) | (nodegree + race) * (scale(age, s=F) + scale(educ, s=F)), data = lalonde, se_type = "HC1") coeftest(fl)["treat",,drop=F]
URI Regression, multi-category treatment
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "race", contrast = c("hispan", "white")) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race == "hispan"]) - weighted.mean(lalonde$re78[lalonde$race != "hispan"], l$weights[lalonde$race != "hispan"]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ relevel(race, "white") + age + educ + married + re74 + re75, data = lalonde) coeftest(f, vcov. = vcovHC)[3,,drop=FALSE]
MRI Regression, multi-category treatment, ATE
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "race", estimand = "ATE") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"]) - weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) ### lmw_est() output summary(lmw_est(l))$coef ### marginaleffects() output f <- lm(re78 ~ race * (age + educ + married + re74 + re75), data = lalonde) summary(marginaleffects(f, var = "race", vcov = vcovHC(f))) ### estimatr output fl <- estimatr::lm_lin(re78 ~ race, ~ age + educ + married + re74 + re75, data = lalonde, se_type = "HC3") summary(fl)$coefficients[1:3,]
MRI Regression, multi-category treatment, ATT
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "race", estimand = "ATT", focal = "black") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race=="hispan"]) - weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"]) ### lmw_est() output summary(lmw_est(l))$coef ### marginaleffects() output f <- lm(re78 ~ race * (age + educ + married + re74 + re75), data = lalonde) summary(marginaleffects(f, var = "race", vcov = vcovHC(f), newdata = subset(lalonde, race == "black")))
URI Regression, multi-category treatment, fixed effects
l <- lmw(re78 ~ race + age + educ + re74 + re75, data = lalonde, method = "URI", treat = "race", contrast = c("hispan", "white"), fixef = ~married) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race == "hispan"]) - weighted.mean(lalonde$re78[lalonde$race != "hispan"], l$weights[lalonde$race != "hispan"]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ relevel(race, "white") + age + educ + married + re74 + re75, data = lalonde) coeftest(f, vcov. = vcovHC)[3,,drop=FALSE] ### estimatr output fl <- estimatr::lm_robust(re78 ~ relevel(race, "white") + age + educ + re74 + re75, data = lalonde, fixed_effects = ~married, se_type = "HC3") coeftest(fl)[2,,drop = FALSE]
Sampling Weights
lalonde$sw <- runif(nrow(lalonde))
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, weights = sw) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATE
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", estimand = "ATE", s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### estimatr output fl <- estimatr::lm_lin(re78 ~ treat, ~ age + educ + race + married + re74 + re75, data = lalonde, se_type = "HC3", weights = sw) summary(fl)$coefficients["treat",,drop=FALSE]
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", fixef = ~race, s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race, data = lalonde, weights = sw) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE] ### lm_robust() output fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, fixed_effects = ~race, se_type = "HC3", weights = sw) coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATE, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", fixef = ~race, s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### URI + covariate version l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race, data = lalonde, method = "URI", treat = "treat", s.weights = sw) weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0]) summary(lmw_est(l2))$coef
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, treat = "treat", iv = ~nodegree + race, s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1"))$coef ### feols() output f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race, data = lalonde, vcov = "HC1", weights = ~sw) coeftest(f)["fit_treat",,drop=F] ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 | nodegree + race + age + educ + married + re74 + re75, data = lalonde, se_type = "HC1", weights = sw) coeftest(fl)["treat",,drop=F]
URI, binary treatment, 2SLS, fixed effects, CR SEs
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, treat = "treat", iv = ~nodegree, fixef = ~race, s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1", cluster = ~race))$coef ### feols() output f <- feols(re78 ~ age + educ + married + re74 + re75 + race | treat ~ nodegree, data = lalonde, cluster = ~race, weights= ~sw) coeftest(f)["fit_treat",,drop=F] ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 | nodegree + race + age + educ + married + re74 + re75, data = lalonde, fixed_effects = ~race, se_type = "stata", cluster = race, weights = sw) coeftest(fl)["treat",,drop=F] #Note: SEs agree for cluster HC1 (aka iv_robust() "stata" and feols() "cluster") and # HC2 (aka iv_robust() "CR2") but not HC0 (aka iv_robust() "CR0"). feols() only # has HC1. Full agreement with vcovCL() used on ivreg::ivreg().
MRI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ, data = lalonde, treat = "treat", method = "MRI", iv = ~nodegree, s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1"))$coef ### iv_robust() age_c_sw <- lalonde$age - weighted.mean(lalonde$age, lalonde$sw) educ_c_sw <- lalonde$educ - weighted.mean(lalonde$educ, lalonde$sw) fl <- estimatr::iv_robust(re78 ~ treat * (age_c_sw + educ_c_sw) | nodegree * (age_c_sw + educ_c_sw), data = lalonde, se_type = "HC1", weights = sw) coeftest(fl)["treat",,drop=F]
URI Regression, multi-category treatment
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "race", contrast = c("hispan", "white"), s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race == "hispan"]) - weighted.mean(lalonde$re78[lalonde$race != "hispan"], l$weights[lalonde$race != "hispan"]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ relevel(race, "white") + age + educ + married + re74 + re75, data = lalonde, weights = sw) coeftest(f, vcov. = vcovHC)[3,,drop=FALSE]
MRI Regression, multi-category treatment, ATE
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "race", estimand = "ATE", s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) - weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"]) ### lmw_est() output summary(lmw_est(l))$coef ### estimatr output fl <- estimatr::lm_lin(re78 ~ race, ~ age + educ + married + re74 + re75, data = lalonde, se_type = "HC3", weights = sw) summary(fl)$coefficients[1:3,]
URI Regression, multi-category treatment, fixed effects
l <- lmw(re78 ~ race + age + educ + re74 + re75, data = lalonde, method = "URI", treat = "race", contrast = c("hispan", "white"), fixef = ~married, s.weights = sw) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "hispan"], l$weights[lalonde$race == "hispan"]) - weighted.mean(lalonde$re78[lalonde$race != "hispan"], l$weights[lalonde$race != "hispan"]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ relevel(race, "white") + age + educ + married + re74 + re75, data = lalonde, weights = sw) coeftest(f, vcov. = vcovHC)[3,,drop=FALSE] ### estimatr output fl <- estimatr::lm_robust(re78 ~ relevel(race, "white") + age + educ + re74 + re75, data = lalonde, fixed_effects = ~married, se_type = "HC3", weights = sw) coeftest(fl)[2,,drop = FALSE]
Base Weights from MatchIt
library(MatchIt) M <- matchit(treat ~ age + educ + race + married + re74 + re75, data = lalonde, method = "nearest", estimand = "ATT") md <- match.data(M, data = lalonde)
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", base.weights = M$weights) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75, data = md, weights = weights) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATT
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", estimand = "ATT", base.weights = M$weights) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### marginaleffects() output f <- lm(re78 ~ treatf * (age + educ + race + married + re74 + re75), data = md, weights = weights) summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f), newdata = subset(md, treat == 1)))
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", fixef = ~race, base.weights = M$weights) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race, data = md, weights = weights) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE] ### lm_robust() output fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75, data = md, fixed_effects = ~race, se_type = "HC3", weights = weights) coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATT, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", fixef = ~race, estimand = "ATT", base.weights = M$weights) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### URI + covariate version l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race, data = lalonde, method = "URI", treat = "treat", estimand = "ATT", base.weights = M$weights) weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0]) summary(lmw_est(l2))$coef ### lm() output f <- lm(re78 ~ treatf * (age + educ + married + re74 + re75) + race, data = md, weights = weights) summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f), newdata = subset(md, treat == 1)))
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, treat = "treat", iv = ~nodegree + race, base.weights = M$weights) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1"))$coef ### feols() output f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race, data = md, vcov = "HC1", weights = ~weights) coeftest(f)["fit_treat",,drop=F] ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 | nodegree + race + age + educ + married + re74 + re75, data = md, se_type = "HC1", weights = weights) coeftest(fl)["treat",,drop=F]
URI, binary treatment, 2SLS, fixed effects, CR SEs
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, treat = "treat", iv = ~nodegree, fixef = ~race, base.weights = M$weights) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1", cluster = ~race))$coef ### feols() output f <- feols(re78 ~ age + educ + married + re74 + re75 + race | treat ~ nodegree, data = md, cluster = ~race, weights = ~weights) coeftest(f)["fit_treat",,drop=F] ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 | nodegree + race + age + educ + married + re74 + re75, data = md, fixed_effects = ~race, se_type = "stata", cluster = race, weights = weights) coeftest(fl)["treat",,drop=F] #Note: SEs agree for cluster HC1 (aka iv_robust() "stata" and feols() "cluster") and # HC2 (aka iv_robust() "CR2") but not HC0 (aka iv_robust() "CR0"). feols() only # has HC1. Full agreement with vcovCL() used on ivreg::ivreg().
Base Weights from WeightIt
library(WeightIt) #Binary treatment, ATE W_ate <- weightit(treat ~ age + educ + race + married + re74 + re75, data = lalonde, method = "ps", estimand = "ATE") #Binary treatment, ATT W_att <- weightit(treat ~ age + educ + race + married + re74 + re75, data = lalonde, method = "ps", estimand = "ATT") #Multi-category treatment, ATE #install.packages("mclogit") W3 <- weightit(race ~ age + educ + married + re74 + re75, data = lalonde, method = "ps", estimand = "ATE", use.mclogit = TRUE, include.obj = TRUE)
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, weights = W_ate$weights) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATE
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### marginaleffects() output f <- lm(re78 ~ treatf * (age + educ + race + married + re74 + re75), data = lalonde, weights = W_ate$weights) summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f)))
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", fixef = ~race, obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race, data = lalonde, weights = W_ate$weights) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE] ### lm_robust() output fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, fixed_effects = ~race, se_type = "HC3", weights = W_ate$weights) coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATE, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", fixef = ~race, obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### URI + covariate version l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race, data = lalonde, method = "URI", treat = "treat", obj = W_ate) weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0]) summary(lmw_est(l2))$coef ### lm() output f <- lm(re78 ~ treatf * (age + educ + married + re74 + re75) + race, data = lalonde, weights = W_ate$weights) summary(marginaleffects(f, var = "treatf", vcov = vcovHC(f)))
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, treat = "treat", iv = ~nodegree + race, obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1"))$coef ### feols() output f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race, data = lalonde, vcov = "HC1", weights = W_ate$weights) coeftest(f)["fit_treat",,drop=F] ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 | nodegree + race + age + educ + married + re74 + re75, data = lalonde, se_type = "HC1", weights = W_ate$weights) coeftest(fl)["treat",,drop=F]
MRI, binary treatment, ATE, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", obj = W_ate, dr.method = "AIPW") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### PSweight() output P <- PSweight::PSweight(ps.estimate = W_ate$ps, zname = "treat", yname = "re78", data = lalonde, weight = "IPW", augmentation = TRUE, out.formula = re78 ~ age + educ + race + married + re74 + re75) summary(P)
MRI, binary treatment, ATT, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", obj = W_att, dr.method = "AIPW") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### PSweight() output P <- PSweight::PSweight(ps.estimate = W_att$ps, zname = "treat", yname = "re78", data = lalonde, weight = "treated", augmentation = TRUE, out.formula = re78 ~ age + educ + race + married + re74 + re75) summary(P) #Note: SEs don't agree for ATT; PSweight's will be a little smaller. See lmw_est.lmw_aipw #for details. PSweight's are more trustworthy because they use a more accurate formula #that involves the propensity score (but it requires a propensity score!). Some #evidence that analytical SEs perform poorly with incorrect outcome model anyway; #Mao, LI, & Greene (2018, SMMR)
URI, binary treatment, ATE, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", obj = W_ate, dr.method = "AIPW") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### PSweight() output fit <- lm(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde) ll0 <- lalonde; ll0$treat <- 0 ll1 <- lalonde; ll1$treat <- 1 y0 <- predict(fit, newdata = ll0) y1 <- predict(fit, newdata = ll1) P <- PSweight::PSweight(ps.estimate = W_ate$ps, zname = "treat", yname = "re78", data = lalonde, weight = "IPW", augmentation = TRUE, out.estimate = cbind(`0` = y0, `1` = y1)) summary(P) #Note: PSweight SE doesn't incorporate estimation of outcome model # because it only supports MRI natively. Point estimates are # correct though.
MRI, multi-category treatment, ATE, AIPW
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "race", obj = W3, dr.method = "AIPW") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) - weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"]) ### lmw_est() output summary(lmw_est(l))$coef ### PSweight() output ps_mat <- fitted(W3$obj); colnames(ps_mat) <- levels(lalonde$race) P <- PSweight::PSweight(ps.estimate = ps_mat, zname = "race", yname = "re78", data = lalonde, weight = "IPW", augmentation = TRUE, out.formula = re78 ~ age + educ + married + re74 + re75) summary(P)
URI, multi-category treatment, ATE, AIPW
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "race", obj = W3, dr.method = "AIPW", contrast = c("white", "black")) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) - weighted.mean(lalonde$re78[lalonde$race != "white"], l$weights[lalonde$race!="white"]) ### lmw_est() output summary(lmw_est(l))$coef ### PSweight() output ps_mat <- fitted(W3$obj); colnames(ps_mat) <- levels(lalonde$race) fit <- lm(re78 ~ race + age + educ + married + re74 + re75, data = lalonde) llb <- lalonde; llb$race[] <- "black" llh <- lalonde; llh$race[] <- "hispan" llw <- lalonde; llw$race[] <- "white" yb <- predict(fit, newdata = llb) yh <- predict(fit, newdata = llh) yw <- predict(fit, newdata = llw) P <- PSweight::PSweight(ps.estimate = ps_mat, zname = "race", yname = "re78", data = lalonde, weight = "IPW", augmentation = TRUE, out.estimate = cbind(black = yb, hispan = yh, white = yw)) summary(P)
Sampling Weights + Base Weights from WeightIt
library(WeightIt) #Binary treatment, ATE W_ate <- weightit(treat ~ age + educ + race + married + re74 + re75, data = lalonde, method = "ps", estimand = "ATE", s.weights = lalonde$sw) #Binary treatment, ATT W_att <- weightit(treat ~ age + educ + race + married + re74 + re75, data = lalonde, method = "ps", estimand = "ATT", s.weights = lalonde$sw) #Multi-category treatment, ATE #install.packages("mclogit") W3 <- weightit(race ~ age + educ + married + re74 + re75, data = lalonde, method = "ps", estimand = "ATE", use.mclogit = TRUE, include.obj = TRUE, s.weights = lalonde$sw)
URI, binary treatment
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", obj = W_ate) #Note: sampling weights and base weights are both in W_ate ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, weights = W_ate$weights * lalonde$sw) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE]
MRI, binary treatment, ATE
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef # Note: not straightforward to compute marginal effects using another package; # need to take weighted average of predicted values where weights are # only the smapling weights, but model is fit using product of sampling # and base weights. marginaleffects() cannot do this. One option is to # center the covariates first at sampling-weighted means.
URI, binary treatment, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", fixef = ~race, obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### lm() output f <- lm(re78 ~ treat + age + educ + married + re74 + re75 + race, data = lalonde, weights = W_ate$weights * lalonde$sw) coeftest(f, vcov. = vcovHC)["treat",,drop = FALSE] ### lm_robust() output fl <- estimatr::lm_robust(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, fixed_effects = ~race, se_type = "HC3", weights = W_ate$weights * lalonde$sw) coeftest(fl)["treat",,drop = FALSE]
MRI, binary treatment, ATE, fixed effects
l <- lmw(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", fixef = ~race, obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef ### URI + covariate version l2 <- lmw(re78 ~ treat * (age + educ + married + re74 + re75) + race, data = lalonde, method = "URI", treat = "treat", obj = W_ate) weighted.mean(lalonde$re78[lalonde$treat == 1], l2$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l2$weights[lalonde$treat==0]) summary(lmw_est(l2))$coef
URI, binary treatment, 2SLS (mutliple IVs)
l <- lmw_iv(re78 ~ treat + age + educ + married + re74 + re75, data = lalonde, treat = "treat", iv = ~nodegree + race, obj = W_ate) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l, robust = "HC1"))$coef ### feols() output f <- feols(re78 ~ age + educ + married + re74 + re75 | treat ~ nodegree + race, data = lalonde, vcov = "HC1", weights = W_ate$weights * lalonde$sw) coeftest(f)["fit_treat",,drop=F] ### iv_robust() fl <- estimatr::iv_robust(re78 ~ treat + age + educ + married + re74 + re75 | nodegree + race + age + educ + married + re74 + re75, data = lalonde, se_type = "HC1", weights = W_ate$weights * lalonde$sw) coeftest(fl)["treat",,drop=F]
MRI, binary treatment, ATE, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", obj = W_ate, dr.method = "AIPW") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef
MRI, binary treatment, ATT, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "MRI", treat = "treat", obj = W_att, dr.method = "AIPW") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef
URI, binary treatment, ATE, AIPW
l <- lmw(re78 ~ treat + age + educ + race + married + re74 + re75, data = lalonde, method = "URI", treat = "treat", obj = W_ate, dr.method = "AIPW") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$treat == 1], l$weights[lalonde$treat==1]) - weighted.mean(lalonde$re78[lalonde$treat == 0], l$weights[lalonde$treat==0]) ### lmw_est() output summary(lmw_est(l))$coef
MRI, multi-category treatment, ATE, AIPW
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "MRI", treat = "race", obj = W3, dr.method = "AIPW") ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) - weighted.mean(lalonde$re78[lalonde$race == "black"], l$weights[lalonde$race=="black"]) ### lmw_est() output summary(lmw_est(l))$coef
URI, multi-category treatment, ATE, AIPW
l <- lmw(re78 ~ race + age + educ + married + re74 + re75, data = lalonde, method = "URI", treat = "race", obj = W3, dr.method = "AIPW", contrast = c("white", "black")) ### Weighted difference in means weighted.mean(lalonde$re78[lalonde$race == "white"], l$weights[lalonde$race=="white"]) - weighted.mean(lalonde$re78[lalonde$race != "white"], l$weights[lalonde$race!="white"]) ### lmw_est() output summary(lmw_est(l))$coef
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.