Fig. 3From: Bayesian Mendelian randomization with an interval causal null hypothesis: ternary decision rules and loss function calibrationLoss of the frequentist (grey flat surfaces) and our Bayesian MR (coloured curves) with binary Y when \(\beta = 0.3\) (outside the tolerance interval). Panels represent different combinations of the missing rate of X (\(80\%,\) \(40\%\) and \(0\%\)) and the IV strength (0.3, 0.1 and 0.05). 200 datasets were simulated for each of these scenariosBack to article page