RD Covariate Balance Test

Tests covariate balance at the RD cutoff by running rdrobust with each covariate as the outcome variable. If the RD design is valid, predetermined covariates should not exhibit a discontinuity at the cutoff.

rd_covariate_balance(
  data,
  covariates,
  x,
  c = 0,
  bw = NULL,
  kernel = "tri",
  p = 1,
  conf_level = 0.95,
  theme_use = causalverse::ama_theme()
)

Arguments

data

A data frame containing the covariates and running variable.

covariates

Character vector. Names of covariate columns to test.

x

Character string. Name of the running variable.

c

Numeric. The cutoff value for the RD design. Default is 0.

bw

Numeric or NULL. Bandwidth to use for all tests. If NULL (the default), rdrobust selects the MSE-optimal bandwidth separately for each covariate.

kernel

Character string. Kernel function for rdrobust. Default is "tri" (triangular).

p

Integer. Order of the local polynomial. Default is 1.

conf_level

Numeric. Confidence level for intervals. Default is 0.95.

theme_use

A ggplot2 theme to apply to the plot. Default is causalverse::ama_theme().

Value

A list with components:

table

A data frame with columns: covariate, estimate, std_error, ci_lower, ci_upper, p_value, bandwidth, n_left, n_right, significant.

plot

A ggplot2 coefficient plot showing the estimate and CI for each covariate.

References

Lee, D. S. and Lemieux, T. (2010). "Regression Discontinuity Designs in Economics." Journal of Economic Literature, 48(2), 281-355.

Examples

if (FALSE) { # \dontrun{
set.seed(42)
n <- 1000
x <- runif(n, -1, 1)
df <- data.frame(
  x   = x,
  age = 30 + 5 * x + rnorm(n),
  inc = 50000 + 10000 * x + rnorm(n, sd = 5000),
  edu = 12 + 2 * x + rnorm(n, sd = 1.5)
)

result <- rd_covariate_balance(
  data       = df,
  covariates = c("age", "inc", "edu"),
  x          = "x",
  c          = 0
)

result$plot
result$table
} # }