local_lm: Local Linear Regression to Estimate Local ATE

In KatoPachi/discreteRD: What the Package Does (One Line, Title Case)

Description

Estimate the local ATE at the treatment assignment threshold by the local linear regression.

Usage

local_lm(
  basemod,
  data,
  subset,
  weights,
  cluster,
  submod,
  order = c(1, 2),
  cutoff,
  assign,
  se = "HC0",
  cholesky = TRUE,
  bw,
  kernel,
  point = 0
)

Arguments

basemod	baseline formula. outcome ~ running variable.
data	data.frame which you want to use.
subset	subset condition.
weights	weight variable.
cluster	cluster variable.
submod	numeric vector. Which baseline model you want to use.
order	numeric vector of polynomial orders.
cutoff	numeric of cutoff points If missing, try to find getOption("discRD.cutoff").
assign	assignment rule of treatment. If "greater", treated whose running variable is greater than or equal to cutoff. If "smaller", treated whose running variable is less than or equal to cutoff. If missing, try to find getOption("discRD.assign").
se	character. How to calculate robust variance-covariance matrix ("HC0", "HC1", "HC2", "HCj", "HC3", and "HC4")
cholesky	logical (default is TRUE). When solving normal equation, use cholesky decomposition.
bw	bandwidth.
kernel	character of kernel density ("uniform" or "triangular")
point	value of the running variable that the kernel weights weigh the most

Examples

running <- sample(1:100, size = 1000, replace = TRUE)
cov1 <- rnorm(1000, sd = 2); cov2 <- rnorm(1000, mean = -1)
y0 <- running + cov1 + cov2 + rnorm(1000, sd = 10)
y1 <- 2 + 1.5 * running + cov1 + cov2 + rnorm(1000, sd = 10)
y <- ifelse(running <= 50, y1, y0)
bin <- ifelse(y > mean(y), 1, 0)
w <- sample(c(1, 0.5), size = 1000, replace = TRUE)
raw <- data.frame(y, bin, running, cov1, cov2, w)

set_optDiscRD(
  y + bin ~ running,
  covmod = list(~cov1, ~ cov1 + cov2),
  discRD.cutoff = 50,
  discRD.assign = "smaller"
)

local <- local_lm(data = raw, bw = 3, kernel = "uniform")
str(local_lm(submod = 1, data = raw, bw = 3, kernel = "uniform"))
str(local_lm(submod = 1, order = 3, data = raw, bw = 3, kernel = "uniform"))
str(local_lm(data = raw, cutoff = 30, bw = 3, kernel = "uniform"))
est2 <- local_lm(
  data = raw, se = "HC1", weights = w, bw = 3, kernel = "uniform"
)

KatoPachi/discreteRD documentation built on Feb. 24, 2022, 12:32 a.m.