We consider four scenarios for the pattern of pleiotropy, and consider the bias and coverage properties of the estimators with null and positive causal effects. The scenarios considered are: no pleiotropy, InSIDE assumption trivially satisfied (all the α parameters, representing the direct effects of genetic variants on the outcome, are equal to zero);balanced pleiotropy, InSIDE assumption satisfied (α parameters take positive and negative values);directional pleiotropy, InSIDE assumption satisfied (α parameters take only positive values, but are generated independently from the γ parameters, representing genetic effects on the exposure);directional pleiotropy, InSIDE assumption not satisfied (α parameters take positive values, and are correlated with the genetic effects on the exposure).
