Fig. 3From: Random forest-based imputation outperforms other methods for imputing LC-MS metabolomics data: a comparative studyHeatmaps. Heat maps representing the average performance of nine imputation methods after 100 permutations; zero (ZERO), ½ -minimum (½ MIN), minimum (MIN), Random forest (RF), mean (MEAN), K-nearest neighbor (KNN), Bayesian Principal Component Analysis (BPCA), Probabilistic Principal Component Analysis (PPCA), Singular Value Decomposition (SVD) in seven missing mechanisms (each box); MCAR, MAR, MNAR, MCAR-MAR, MCAR-MNAR, MAR-MNAR, MCAR-MAR-MNAR. The darker blue color indicates that the error is small whereas, when the color transforms to lighter shades, this is an indication that the error becomes higher by the corresponding imputation methodBack to article page