This post was inspired by this paper of Karsten Luebke and coauthors. We’ll stratify our sample into two groups: students (Studium) and non-students (kein Studium). Structural causal model First, we define the structure of our causal model. set.seed(42) # reproducibilty N <- 1e03 IQ = rnorm(N) Fleiss = rnorm(N) Eignung = 12 * IQ + 12 * Fleiss + rnorm(N, 0, .1) That is, aptitude (Eignung) is a function of intelligence (IQ) and dilligence (Fleiss), where the input variables have the same impact on the outcome variable …