In ytalau/eiv: The Study of Measurement Errors in Variables

knitr::opts_chunk$set(comment = "", cache = TRUE)
options(width = 100, digits = 5)

The R package eiv provides a main function eivgmm to correct measurement error in a longitudinal setting.

Installation

You can install the development version at GitHub using the follwing code,

if (! require(devtools)) install.packages("devtools")
devtools::install_github("https://github.com/ytalau/eiv")

Supported families

Gaussian (identity link)
Binomial (logit link)

Example

Linear regression:

library(eiv)
set.seed(8593)
n <- 200 # number of independent clusters
m <- 5 # cluster size
X <- rnorm(n * m, mean = 0, sd = 1)
W <- X + rnorm(n * m, mean = 0, sd = 0.5)
Y <- X * 1 + rnorm(n * m, 0, 1)
dat <- data.frame(X = X, W = W, Y = Y, time = rep(1:m, n), 
                      id = rep(1:n, each = m))
init <- glm(Y ~ W, data = dat)
coef(init)

cres <- eiv::eivgmm(Y ~ W ,
                    data = dat, me.var = "W",
                    mcov = lapply(1:m, function(i) as.matrix(0.25)),
                    time.var = "time", id.var = "id",
                    family = "gaussian",
                    init.beta = init$coefficients,
                    modify_inv = 0,
                    finsam_cor = 0)
coef(cres)