Talent and Looks -- Collider bias

Background Some musing on the collider bias. Let’s try to reverse engineer this image Setup library(tidyverse) library(ggdag) Simulate some data n <- 1000 d <- tibble( x = rnorm(n, mean = 0, sd = 1), y = rnorm(n, mean = 0, sd = 1), e = rnorm(n, mean = 0, sd = 0.3), z = abs(x) * abs(y)) d: Uncorrelated data The farer from the centroid the lighter the …