Skip to main content
International Journal of Epidemiology logoLink to International Journal of Epidemiology
. 2021 Jan 19;50(2):699. doi: 10.1093/ije/dyab006

Educational note: Paradoxical collider effect in the analysis of non-communicable disease epidemiological data: a reproducible illustration and web application

Miguel Angel Luque-Fernandez, Michael Schomaker, Daniel Redondo-Sanchez, Maria Jose Sanchez Perez, Anand Vaidya, Mireille E Schnitzer
PMCID: PMC8128475  PMID: 33462617

First published online: 10 December 2019, Int J Epidemiol 2019; Volume 48, Issue 2, April 2019, Pages 640–653. doi: https://doi.org/10.1093/ije/dyy275

The corrigendum to this article published on 10 December 2019 (https://doi.org/10.1093/ije/dyz247) referred to corrections made to the code in Box 2. These corrections were in fact, incorrect, as was the sentence referring to Box 2. The correct text and Box 2 code is as follows (changes shown in bold).

Box 2. To generate data consistent with Figure 2B

library(visreg) # load package to visualize regression output

library(ggplot2) # load package to visualize regression output

N <- 1000 # sample size

set.seed(777)

A <- rnorm(N) # exposure

Y <- 0.3 * A + rnorm(N)

C <- 1.2 * A + 0.9 * Y + rnorm(N)

fit3 <- lm(Y ∼ A) # crude model

fit4 <- lm(Y ∼ A + C) # adjusted model

# visualize adjusted model

g2 < - visreg(fit4, ‘A’, gg = TRUE, line = list(col = ‘red’),

points = list(size = 2, pch = 1, col = ‘black’)) + theme_classic()+

coord_cartesian(ylim = c(-4, 4)) +

ggtitle(“Figure 2B”)

“The true causal coefficient of the exposure A is 0.3, and the coefficients for the association of the collider C with the exposure A and the outcome Y are 1.2 and 0.9, respectively (Box 2).”


Articles from International Journal of Epidemiology are provided here courtesy of Oxford University Press

RESOURCES