Mechanism-aware imputation: a two-step approach in handling missing values in metabolomics

BMC Bioinformatics

Table 2 Accuracies (across 450 simulation repetitions) of the random forest classifier in step one of MAI

	Total % missing	49.5% imposed MCAR	60% imposed MCAR	80% imposed MCAR	90% imposed MCAR
Mean accuracy [95% CI]	10%	98% [97%, 98%]	96% [96%, 97%]	93% [92%, 94%]	92% [91%, 93%]
Mean accuracy [95% CI]	30%	92% [91%, 93%]	88% [87%, 89%]	82% [80%, 84%]	90% [88%, 91%]
Median accuracy [95% CI]	10%	98% [97%, 98%]	97% [96%, 97%]	93% [92%, 94%]	92% [91%, 93%]
Median accuracy [95% CI]	30%	92% [91%, 93%]	88% [87%, 89%]	82% [80%, 84%]	90% [88%, 92%]

Mean (two top rows) and median accuracies (two bottom rows) are reported for different levels of total percent missing and percent imposed MCAR missingness on the COPDGene Data Set 1; p = 300 and n = 100

ISSN: 1471-2105