The performance evaluation of 8 different MVI methods on the OC dataset. The figure is divided into 4 blocks, each block representing a different missingness range from 10% to 40%, with missingness being combined of MNAR and MCAR. The x-axis in each block indicates the proportion of MCAR within the combination. The three subfigures within each block provide different perspectives on the performance of the MVI methods. A) The NRMSE score across all combinations, with a lower score indicating a smaller error. For better comparison, the y-axis is labeled from 1.00 at the bottom to 0.00 at the top, so that the better-performing methods are displayed at the top of the chart. B) The rank of Procrustes SS across all combinations, with a smaller rank indicating better performance. The rank is numbered from 1 to 8. C) The heatmap of the rank of the Correlation coefficient, with the darker color indicating a smaller rank. The Correlation coefficient measures the similarity between two datasets, and a smaller rank indicates a better performance of the MVI method in terms of preserving the original relationship between the variables in the data.