Robust Significance Analysis of Microarrays by Minimum β-Divergence Method
Table 3
Performance evaluation of different methods using AUC, pAUC, and FDR values for both small- and large-sample cases.
Performance measures
ANOVA
KW
SAM
LIMMA
Proposed
For small-sample case
AUC
0.764
0.196
0.832
0.834
0.832
(0.279)
(0.102)
(0.192)
(0.279)
(0.839)
[0.084]
[0.009]
[0.177]
[0.097]
[0.839]
pAUC
0.152
0.038
0.166
0.166
0.166
(0.055)
(0.019)
(0.038)
(0.055)
(0.167)
[0.016]
[0.019]
[0.035]
[0.019]
[0.167]
FDR
0.235
0.802
0.167
0.165
0.167
(0.720)
(0.897)
(0.807)
(0.720)
(0.160)
[0.915]
[0.900]
[0.822]
[0.902]
[0.160]
For large-sample case
AUC
0.957
0.957
0.957
0.957
0.957
(0.446)
(0.864)
(0.594)
(0.446)
(0.937)
[0.227]
[0.857]
[0.487]
[0.227]
[0.947]
pAUC
0.191
0.191
0.191
0.191
0.191
(0.088)
(0.172)
(0.118)
(0.088)
(0.187)
[0.045]
[0.171]
[0.097]
[0.045]
[0.188]
FDR
0.042
0.042
0.045
0.042
0.042
(0.552)
(0.135)
(0.405)
(0.552)
(0.062)
[0.772]
[0.145]
[0.512]
[0.772]
[0.057]
In this table the values within the brackets (⋅), , and [⋅] represent the summary statistics estimated by different methods (ANOVA, KW, SAM, LIMMA, and proposed) in presence of one or two outliers in each of 10%, 20%, and 50% genes, respectively.