Figure 1From: Tuning multiple imputation by predictive mean matching and local residual draws Bivariate plots are of x vs. z values in a single imputed dataset. Observed x in purple circles; imputed in blue crosses. Left to right: normal errors parametric imputation, PMM and LRD (type 1 matching with k=3). Top to bottom: Non-normal residuals, heteroscedasticity and non-linearity. These scenarios represent problems for a linear normal errors imputation model.Back to article page