Fig. 1
From: Introduction to instrumental variables and their application to large-scale assessment data
Examples of a situation where the modeling of causal relationships using observational data will be biased (a) and a situation where it will be valid (b)