Table A7.
Imputation Method | Dataset Sizes | ||
---|---|---|---|
n = 599 | n = 5400 | n = 20,000 | |
MICE-pmm (minutes) | 1.9 | 13.4 | 138.9 |
kNN-obs-sel (minutes) | 0.7 | 16.2 | 210.7 |
Imputation was applied to the actual NEO dataset (n = 599; 58% metabolites contain missing values) and two oversampled datasets generated from the NEO data. With n = 599, kNN-obs-sel was slightly faster. However, MICE-pmm imputation took a shorter time to complete the imputations in larger datasets.