In order to investigate the performance of the weighted median method in realistic settings, as well as to determine in what scenarios it performs well or badly in comparison with the IVW and MR‐Egger regression methods, we perform a simulation study. We assume there are 25 genetic variants that are candidate IVs, and consider three scenarios: 1.Balanced pleiotropy, InSIDE assumption satisfied—pleiotropic effects are equally likely to be positive as negative, these effects are uncorrelated with the instrument strength.2.Directional pleiotropy, InSIDE assumption satisfied—only positive pleiotropic effects are simulated, these effects are uncorrelated with the instrument strength.3.Directional pleiotropy, InSIDE assumption not satisfied—pleiotropic effects are via a confounder, these effects on the outcome are therefore positive and are correlated with the instrument strength.
