NonlinearRDD (version 0.0.4)

NRDD: Nonlinear RDD estimator

Description

Semiparametric estimation of the possibly nonlinear and nonseparable structural function in regression discontinuity designs with a continuous treatment variable

Usage

NRDD(
  df,
  b1 = 2 * nrow(df)^{
     -1/5
 },
  b2 = 2 * nrow(df)^{
     -1/5
 },
  b3 = 2 * nrow(df)^{
     -1/5
 },
  t_tilde = stats::median(df[, 2]),
  e_space,
  Nonseparable = TRUE
)

Value

The estimated parameters in the structural function.

Arguments

df

Data frame whose columns are outcome Y, treatment Treat, and running variable R, respectively.

b1

Bandwidth for the local linear estimation of the conditional distribution of Y given T,R. Default is n^(-1/5), where n is the sample size.

b2

Bandwidth for the smoothing of the indicator function 1(Y <= y).

b3

Bandwidth for the first-step conditional quantile estimation of T given R.

t_tilde

Normalization constant in the structural function.

e_space

The space of epsilon on which the numerical integration is conducted.

Nonseparable

Whether the structural function should be estimated as a separable or nonseparable function of the treatment. The default is TRUE.

Examples

n <- 20
set.seed(123)
Treat <- runif(n)
R <- runif(n)-0.5
epsilon <- Treat^2 + R
Y <- Treat + R + epsilon
df <- data.frame(Y=Y,Treat=Treat,R=R)
NRDD(df,e_space = seq(0,1,by=0.5),Nonseparable=FALSE)

# \donttest{
n <- 500
set.seed(123)
data <- nrddgp(n)
NRDD(data,e_space = seq(-10,10,by=1))
# }