Skip to main content

Table 2 Average RMSE of 100 simulations using the human atherothrombotic dataset for KNN-TN, KNN-CR and KNN-EU

From: Distribution based nearest neighbor imputation for truncated high dimensional data with applications to pre-clinical and clinical metabolomics studies

MNAR/MAR

Sample size

Group

KNN-TN

KNN-CR

KNN-EU

6%/3%

50

sCAD

1.145 (0.047)

1.171 (0.046)

1.410 (0.052)

50

TYPE1

1.255 (0.054)

1.273 (0.053)

1.555 (0.057)

50

TYPE2

1.266 (0.051)

1.279 (0.050)

1.567 (0.055)

100

sCAD

1.083 (0.048)

1.109 (0.041)

1.403 (0.053)

100

TYPE1

1.183 (0.048)

1.199 (0.041)

1.531 (0.053)

100

TYPE2

1.183 (0.048)

1.191 (0.041)

1.531 (0.053)

10%/5%

50

sCAD

1.146 (0.045)

1.168 (0.045)

1.337 (0.050)

50

TYPE1

1.262 (0.059)

1.280 (0.057)

1.490 (0.059)

50

TYPE2

1.296 (0.048)

1.315 (0.047)

1.531 (0.051)

100

sCAD

1.075 (0.031)

1.095 (0.031)

1.330 (0.034)

100

TYPE1

1.171 (0.039)

1.189 (0.038)

1.460 (0.041)

100

TYPE2

1.189 (0.040)

1.207 (0.038)

1.490 (0.040)

20%/10%

50

sCAD

1.120 (0.049)

1.140 (0.049)

1.210 (0.047)

50

TYPE1

1.261 (0.061)

1.282 (0.061)

1.398 (0.059)

50

TYPE2

1.354 (0.058)

1.373 (0.058)

1.484 (0.054)

100

sCAD

1.033 (0.035)

1.053 (0.035)

1.198 (0.034)

100

TYPE1

1.153 (0.041)

1.176 (0.041)

1.372 (0.041)

100

TYPE2

1.246 (0.037)

1.266 (0.037)

1.451 (0.036)

  1. Total missing was considered at 9%, 15% and 30%, and within each missing, MNAR was greater than MAR