runSimulation: Run a simulation with a specified caliper width, risk...

In kkbrum/PSMOptimalCalipers: Reproduces and Extends Simulations on Optimal Caliper Width in Propensity-Score Matching

Description

This runs a simulation that creates N data sets, each of size 10,000, that have a specified risk difference and covariate distribution. Propensity score matching with a caliper derived from the one given gamma is then conducted and the risk difference in the full and matched data sets is estimated. The average estimated risk difference for the full and matched data sets is reported, as well as the bias reduction of matching and the MSE of the matched estimator. The mean and variance of the time it took to generate 1 data set, match, and calculate risk differences is also reported.

Usage

runSimulation(
  gamma,
  N,
  true_risk_diff,
  cov_dist,
  alpha_0_treat,
  alpha_0_outcome,
  beta
)

Arguments

gamma	The caliper width in terms of pooled standard deviations
N	The number of data sets to generate within the simulation
true_risk_diff	The desired treated risk minus control risk
cov_dist	The covariate distribution. An integer 1-5 or one of the following: 'Ind Norm', 'Corr Norm', '5 Bern', '9 Bern', '10 Bern'
alpha_0_treat	The calculated value of alpha_0_treat, obtained from calcAlpha0()
alpha_0_outcome	The calculated value alpha_0_outcome, obtained from calcAlpha0()
beta	The calculated value of beta, obtained from calcBeta()

Value

A named vector containing the gamma value used, MSE of the matched estimator, the bias reduction of using the matched estimator, the average estimate from the full data, the average estimate from the matched data, the average time to go through one data set, and the variance of the times to go through one data set.

kkbrum/PSMOptimalCalipers documentation built on Dec. 21, 2021, 6:45 a.m.