Robust Significance Analysis of Microarrays by Minimum β-Divergence Method
Table 5
Performance evaluation of different methods based on Spike gene expression dataset for randomly chosen subsamples of size 3 from each condition.
Methods
TPR
TNR
FPR
FNR
FDR
AUC
In absence of outliers
ANOVA
0.787
0.975
0.024
0.213
0.213
0.784
KW
0.788
0.975
0.024
0.211
0.219
0.781
SAM
0.821
0.979
0.020
0.178
0.175
0.820
LIMMA
0.816
0.978
0.021
0.183
0.186
0.814
EB
0.644
0.958
0.041
0.356
0.356
0.642
GAGA
0.598
0.953
0.046
0.401
0.407
0.595
BRIDGE
0.644
0.958
0.041
0.355
0.355
0.642
Proposed
0.821
0.978
0.021
0.185
0.175
0.820
In presence of one outlier in 20% of 1944 DE genes
ANOVA
0.630
0.957
0.043
0.370
0.370
0.622
KW
0.620
0.955
0.044
0.380
0.380
0.611
SAM
0.730
0.968
0.031
0.270
0.270
0.726
LIMMA
0.720
0.967
0.032
0.284
0.284
0.712
EB
0.114
0.897
0.102
0.885
0.885
0.111
GAGA
0.608
0.954
0.045
0.392
0.392
0.605
BRIDGE
0.668
0.961
0.038
0.331
0.331
0.665
Proposed
0.806
0.977
0.022
0.193
0.193
0.805
In this table the summary statistics (TPR, TNR, FPR, FNR, FDR, and AUC) were estimated by different methods (ANOVA, KW, SAM, LIMMA, EB, GAGA, BRIDGE, and proposed) based on valid 1944 DE genes in both absence and presence of one outlier in 20% of valid 1944 DE genes.