Skip to main content

Table 4 MNAR:low missing values

From: Methylation data imputation performances under different representations and missingness patterns

 

MAE

RMSE

Method

M-value

B-value

M-value

B-value

(a) Healthy datasets

    

mean

0.022 ±0.001

0.023 ±0.001

0.043 ±0.001

0.044 ±0.001

impute.knn

0.041 ±0.012

0.033 ±0.006

0.086 ±0.021

0.077 ±0.014

softImpute

0.026 ±0.002

0.023 ±0.003

0.052 ±0.006

0.046 ±0.010

imputePCA

0.019 ±0.001

0.020 ±0.001

0.039 ±0.002

0.039 ±0.001

SVDmiss

0.029 ±0.001

0.021 ±0.001

0.061 ±0.003

0.041 ±0.002

missForest

0.019 ±0.001

0.020 ±0.001

0.037 ±0.002

0.038 ±0.001

methyLImp

0.022 ±0.001

0.019 ±0.001

0.040 ±0.002

0.039 ±0.002

(b) Disease datasets

    

mean

0.036 ±0.001

0.037 ±0.001

0.068 ±0.002

0.069 ±0.002

impute.knn

0.063 ±0.014

0.048 ±0.008

0.120 ±0.020

0.102 ±0.015

softImpute

0.040 ±0.005

0.036 ±0.004

0.076 ±0.010

0.072 ±0.013

imputePCA

0.031 ±0.001

0.032 ±0.001

0.061 ±0.002

0.062 ±0.002

SVDmiss

0.047 ±0.001

0.035 ±0.001

0.089 ±0.003

0.070 ±0.003

missForest

0.031 ±0.001

0.032 ±0.001

0.060 ±0.002

0.061 ±0.002

methyLImp

0.032 ±0.001

0.028 ±0.001

0.063 ±0.003

0.058 ±0.002

  1. Average Mean Average Error (MAE) and Root Mean Square Error (RMSE) imputation performance ± standard deviation. For each method, the symbol indicates the measure (either M-value or β-value) for which the Wilcoxon signed-rank test p-value is <0.05. Best results per metric with respect to the Wilcoxon signed-rank test are highlighted in bold