Fig. 1
From: Evaluation of a weighting approach for performing sensitivity analysis after multiple imputation
Graphical diagnostics (n = 500, m = 300, δ(True) = 0.2, β(True) = 0.5, \( {\widehat{\beta}}_{(Simulate)} \) =0.501). Left panel: the normalised weight (calculated using Equation (7)) for each of the 300 imputed datasets, plotted against the estimated regression coefficient of Y on X obtained under MAR. The vertical dashed line corresponds to the pooled MI estimate under MAR. Right panel: the running mean of the MNAR estimate of the regression coefficient of Y on X, plotted against the number of imputed datasets. The estimates of the same coefficient obtained under MAR are plotted on the right y-axis for each of the 300 imputed datasets. The horizontal dashed line represents the pooled MI estimate under MAR