samorso / IBpaper

This repository contains the code to replicate the numerical studies presented in the paper "A Flexible Bias Correction Method based on Inconsistent Estimators".

IBpaper

+ inside

Overview

This package contains the code we use for the simulations and real data analysis in the paper Zhang et al. (2022).

To install this package:

## if not installed
## install.packages("remotes")
remotes::install_github("samorso/IBpaper")

We highly rely on the ib package (v0.2.0):

install.packages("ib")

You also need the following packages:

install.packages(c("betareg", "BH", "MASS", "Rcpp", "RcppEigen", "RcppNumerical"))

Usage

Here is a simple example to run a simulation for a logistic regression:

library(IBpaper)
library(ib)

set.seed(6)
x <- matrix(rnorm(300), ncol = 3) # design matrix
beta <- 1:4 # regression coefficients
logistic_object <- make_logistic(x, beta) # see ?`make_logistic`
y <- simulation(logistic_object) # from `ib` package

fit_mle <- glm(y ~ x, family = binomial(link = "logit")) # fit logistic regression
fit_jini <- ib(fit_mle, control=list(H=200, verbose=TRUE)) # iterative bootstrap procedure from `ib` package
results <- data.frame(MLE = coef(fit_mle), JINI = coef(fit_jini), "True parameter" = beta, check.names = FALSE)
results

            MLE  JINI True parameter
(Intercept) 1.06 0.87 1.00
x1          2.74 2.34 2.00
x2          3.42 2.95 3.00
x3          4.52 3.89 4.00

The detailed code to run the simulations and real data analysis in the paper Zhang et al. (2022) can be found in this link.