Fig. 1: Comparison of the methods based on mean squared error criterion.
From: A robust and efficient method for Mendelian randomization with hundreds of genetic variants
The mean squared error of the various methods is plotted in each scenario with a null causal effect. The corresponding plot with a positive causal effect was practically identical. The contamination mixture method has the best overall performance according to this measure, particularly in scenarios where 40 or 60 out of the 100 genetic variants were invalid instruments. The vertical axis is plotted on a logarithmic scale.
