Simulate an experimental dataset with selection bias

Usage

selDGP(
  n = 10000,
  p = 10,
  treatProb = 0.5,
  xb = c(-1, 1),
  piF = function(x) 1/2,
  tauF = function(x) 1/exp(-x[3]),
  y0F = function(x) 3 * pmax(x[1] + x[2], 0) + 5 * sin(x[5]) * 2 * pmax(x[7], 0.5),
  selF = function(x) x[1] - 5 * x[3] + pmax(x[4], 0)
)

Arguments

n: sample size (default 10000)
p: number of covariates (default 10)
xb: bounds of uniform random variables for X (default -1, 1)
piF: propensity score function (fixed number indicates simple randomization - default 1/2)
tauF: treatment heterogeneity function (nonlinear by default)
y0F: baseline outcome function (nonlinear by default)
selF: selection bias function (nonlinear by default). When null, no missingness is introduced.

Value

list with outcome (with missing values corresponding with s = 0), treatment, covariates, selection, and true treatment effect