Fig. 4From: Distribution based nearest neighbor imputation for truncated high dimensional data with applications to pre-clinical and clinical metabolomics studiesBoxplots of root mean squared error for KNN-TN, KNN-CR and KNN-EU for 100 datasets, 50 samples by 400 metabolites. Total missing was considered at 9%, 15% and 30% and within each missing MNAR is greater than MAR. The three correlation structure used was i) only positive correlation 0.7, ii) AR(1) correlation 0.9 and iii) mixed correlation 0.7Back to article page