Fig. 5
From: Evaluation of a weighting approach for performing sensitivity analysis after multiple imputation
Graphical procedure for selecting a range for δ (top). Top left panel: histogram of the sum of imputed Y values for each dataset. Top right panel: graphical determination of delta for the variable Y using the approach suggested by Héraud-Bousquet et al. [30]. Graphical diagnostics(bottom). Bottom left panel: normalised weights against the MAR estimates obtained from each of the 300 imputed dataset. Bottom right panel: mean of the MNAR estimate against the number of imputations (n = 500, m = 300, δ(True) = 1, β(True) = 0.5, \( {\widehat{\beta}}_{(Simulate)} \) =0.501)