attcpo: attcpo

In csabounds: Bounds on Distributional Treatment Effect Parameters

Description

compute the Average Treatment Effect on the Treated Conditional on the previous outcome (ATT-CPO)

Usage

attcpo(formla, t, tmin1, tmin2, tname, data, idname, Y0tqteobj, h = NULL,
  yseq = NULL, yseqlen = 100, se = TRUE, iters = 100,
  method = "level")

Arguments

formla	e.g. y ~ treat
t	the last time period
tmin1	the middle time period
tmin2	the first time period
tname	the name of the column containing time periods in the data
data	a data.frame
idname	the name of the column containing an individual identifier over time
Y0tqteobj	a qte object (from the qte package) containing the the counterfactual distribution of untreated potential outcomes for the treated group
h	optional bandwidth
yseq	optional sequence of y values, default is to use all unique yvalues in the data, though this can increase computation time
yseqlen	optional length of y values to use, aids in automatically generating yseq if desired
se	whether or not to compute standard errors
iters	how many bootstrap iterations to use if computing standard errors; default is 100.
method	should be either "levels" or "rank"; whether to compute the ATT-CPO using based on the levels of Y0tmin1 or the ranks of Y0tmin1; "levels" is the default.

Value

att-cpo

Examples

data(displacements)
cc <- qte::CiC(learn ~ treat,
               t=2011, tmin1=2007, tname="year",
               idname="id", panel=TRUE, data=displacements,
               probs=seq(.05,.95,.01),se=FALSE)
cc$F.treated.tmin1 <- ecdf(subset(displacements, year==2007 & treat==1)$learn)
cc$F.treated.tmin2 <- ecdf(subset(displacements, year==2003 & treat==1)$learn)
ac <- attcpo(learn ~ treat, 2011, 2007, 2003, "year", displacements,
        "id", cc, method="rank", yseqlen=10)
ac
ggattcpo(ac)

csabounds documentation built on May 1, 2019, 8:47 p.m.