kmldte: Kaplan-Meier Local Distributional Treatment Effect
In pedrohcgs/kmte: An R Package for Treatment Effects with Censored Outcomes

kmldte computes the Local Distributional Treatment Effect for possibly right-censored outcomes. The estimator relies on the availability of an Instrumental variable Z, and on a monotonicity assumption. To implement the estimator, we make use of an instrumental propensity score approach. For details of the estimation procedure, see Sant'Anna (2016a), 'Program Evaluation with Right-Censored Data'.

1 2	kmldte(out, delta, treat, z, xpscore, ysup = NULL, b = 1000, ci = c(0.9, 0.95, 0.99), cores = 1, monot = TRUE)

`out`	vector containing the outcome of interest
`delta`	vector containing the censoring indicator (1 if observed, 0 if censored)
`treat`	vector containing the treatment indicator (1 if treated, 0 if control)
`z`	vector containing the binary instrument
`xpscore`	matrix (or data frame) containing the covariates (and their transformations) to be included in the instrument propensity score estimation. Instrument Propensity score estimation is based on Logit.
`ysup`	scalar or vector of points for which the distributional treatment effect is computed. If NULL, all uncensored data points available are used.
`b`	The number of bootstrap replicates to be performed. Default is 1,000.
`ci`	A scalar or vector with values in (0,1) containing the confidence level(s) of the required interval(s). Default is a vector with 0,90, 0.95 and 0.99
`cores`	number of processesors to be used during the bootstrap (default is 1). If cores>1, the bootstrap is conducted using snow.
`monot`	Default is TRUE, which impose that the estimated counterfactual distributions are in proper CDF's, i.e. takes values between [0,1], and are non-decreasing. Boundedness is imposed by truncantion, and monotonicity is imposed using the rearrangement procedure proposed by Chernozhukov, Fernandez-Val, and Galichon (2010), implemented in R through package Rearrangement. If FALSE, no adjustment is made.

a list containing the local distributional treatment effect estimate, ldte, and the bootstrapped ci confidence confidence interval, ldte.lb (lower bound), and ldte.ub (upper bound).

pedrohcgs/kmte documentation built on May 24, 2019, 11:46 p.m.