kernelDiscreteDensitySmooth: Density and/or kernel regression estimator with conditioning...

In smoothemplik: Smoothed Empirical Likelihood

Density and/or kernel regression estimator with conditioning on discrete variables

Description

Density and/or kernel regression estimator with conditioning on discrete variables

Usage

kernelDiscreteDensitySmooth(x, y = NULL, compact = FALSE, fun = mean)

Arguments

x	A vector or a matrix/data frame of discrete explanatory variables (exogenous). Non-integer values are fine because the data are split into bins defined by interactions of these variables.
y	Optional: a vector of dependent variable values.
compact	Logical: return unique values instead of full data with repeated observations?
fun	A function that computes a statistic of y inside every category defined by x.

Value

A list with x, density estimator (fhat) and, if y was provided, regression estimate.

Examples

set.seed(1)
x <- sort(rnorm(1000))
p <- 0.5*pnorm(x) + 0.25 # Propensity score
d <- as.numeric(runif(1000) < p)
# g = discrete version of x for binning
g <- as.numeric(as.character(cut(x, -4:4, labels = -4:3+0.5)))
dhat.x <- kernelSmooth(x = x, y = d, bw = 0.4, no.dedup = TRUE)
dhat.g <- kernelDiscreteDensitySmooth(x = g, y = d)
dhat.comp <- kernelDiscreteDensitySmooth(g, d, compact = TRUE)
plot(x, p, ylim = c(0, 1), bty = "n", type = "l", lty = 2)
points(x, dhat.x, col = "#00000044")
points(dhat.comp, col = 2, pch = 16, cex = 2)
lines(dhat.comp$x, dhat.comp$fhat, col = 4, pch = 16, lty = 3)

smoothemplik documentation built on Aug. 22, 2025, 1:11 a.m.