Skip to main content

Table 3 Test AUC for four real data sets: each predictor is constructed by the DLDA rule. These values indicate mean(sd) and were evaluated by 100 random separations of the full data

From: Reproducible detection of disease-associated markers from gene expression data

breast cancer data

AUC of the test data by DLDA

 

t

t 1,1

s 1,1

s 1,5

s 1,10

10 genes

0.698 (0.058)

0.698 (0.058)

0.698 (0.060)

0.684 (0.065)

0.679 (0.069)

50 genes

0.705 (0.046)

0.705 (0.047)

0.707 (0.050)

0.712 (0.051)

0.712 (0.050)

100 genes

0.711 (0.045)

0.710 (0.045)

0.712 (0.047)

0.718 (0.045)

0.718 (0.045)

Cohort data

AUC of the test data by DLDA

 

t

t 1,1

s 1,1

s 1,5

s 1,10

10 genes

0.744 (0.061)

0.743 (0.061)

0.751 (0.064)

0.755 (0.064)

0.771 (0.063)

50 genes

0.779 (0.057)

0.778 (0.057)

0.773 (0.053)

0.784 (0.053)

0.789 (0.054)

100 genes

0.782 (0.056)

0.781 (0.057)

0.778 (0.054)

0.781 (0.052)

0.784 (0.051)

Prostate cancer data

AUC of the test data by DLDA

 

t

t 1,1

s 1,1

s 1,5

s 1,10

10 genes

0.835 (0.025)

0.835 (0.026)

0.832 (0.025)

0.823 (0.027)

0.808 (0.032)

50 genes

0.846 (0.023)

0.845 (0.024)

0.847 (0.022)

0.836 (0.029)

0.829 (0.033)

100 genes

0.844 (0.024)

0.842 (0.025)

0.848 (0.021)

0.829 (0.030)

0.822 (0.033)

Breast cancer data2

AUC of the test data by DLDA

 

t

t 1,1

s 1,1

s 1,5

s 1,10

10 genes

0.612 (0.044)

0.614 (0.041)

0.611 (0.043)

0.595 (0.042)

0.581 (0.044)

50 genes

0.634 (0.040)

0.633 (0.040)

0.630 (0.042)

0.623 (0.039)

0.619 (0.040)

100 genes

0.637 (0.040)

0.636 (0.038)

0.636 (0.416)

0.630 (0.037)

0.626 (0.038)

Leukemia data

AUC of the test data by DLDA

 

t

t 1,1

s 1,1

s 1,5

s 1,10

10 genes

0.981 (0.017)

0.982 (0.017)

0.986 (0.014)

0.991 (0.012)

0.990 (0.016)

50 genes

0.992 (0.014)

0.992 (0.014)

0.988 (0.013)

0.994 (0.008)

0.994 (0.008)

100 genes

0.992 (0.016)

0.991 (0.017)

0.989 (0.013)

0.995 (0.006)

0.995 (0.009)