In each case, we impose a fourth identifying assumption of homogeneity, meaning that the causal effect does not vary across levels of a single instrument, nor across instruments. For all analysis a two-stage regression approach was used. In the continuous smoking exposure case, the smoking variab...