Similar to IVW but uses a simulation-based approach to detect outlying variants and flags them for removal in order to re-estimate the original exposure-outcome relationship. It has been shown that this outlier detection framework is closely related to the analytical approach of applying IVW with so-called “modified second order weights”.
Outlier removal methods can effectively reduce bias in MR estimates, but caution must be made because they will necessarily reduce the standard error and could lead to over-fitting. Other outlier removal methods have also been implemented elsewhere in GSMR and Radial MR.
- Bowden J, Spiller W, Del Greco M F et al. Improving the visualization, interpretation and analysis of two- sample summary data Mendelian randomization via the Radial plot and Radial regression. International Journal of Epidemiology 2018;47:1264-1278.