Comparison fits of the multiplicative structural mean model

In OneSampleMR: One Sample Mendelian Randomization and Instrumental Variable Analyses

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(OneSampleMR)

Comparison fits

• Example from the ivtools ivglm() helpfile. First simulate some example data.

set.seed(12345)
n    <- 5000
psi0 <- 0.5
psi1 <- 0.2
Z    <- rbinom(n, 1, 0.5)
X    <- rbinom(n, 1, 0.7 * Z + 0.2 * (1 - Z))
m0   <- plogis(1 + 0.8 * X - 0.39 * Z)
Y    <- rbinom(n, 1, plogis(psi0 * X + log(m0 / (1 - m0))))
dat  <- data.frame(Z, X, Y)

• Comparison fit using msmm().

fit02 <- msmm(Y ~ X | Z, data = dat)
summary(fit02)

• Comparison fit using the alternative GMM moment conditions.

fit03 <- msmm(Y ~ X | Z, data = dat, estmethod = "gmmalt")
summary(fit03)

• Comparison fit using transformed variables using two-stage least squares.

fit04 <- msmm(Y ~ X | Z, data = dat, estmethod = "tsls")
summary(fit04)

• Comparison fit using the alternative transformed variables approach using two-stage least squares.

fit05 <- msmm(Y ~ X | Z, data = dat, estmethod = "tslsalt")
summary(fit05)

OneSampleMR documentation built on June 8, 2025, 10:51 a.m.