Skip to main content

Table 2 P-values comparing FDR by different methods.

From: Re-sampling strategy to improve the estimation of number of null hypotheses in FDR control under strong correlation structures

 

ρ

True π0

BY

SampG.q

SampS.q

SampG.m

SampS.m

Data-B

0.3

0.9

3.25E-11

3.68E-01

1.56E-02

1.31E-01

1.29E-01

  

0.8

7.45E-26

8.70E-04

8.40E-04

7.20E-03

4.12E-03

  

0.7

2.80E-32

6.23E-07

1.34E-07

4.13E-06

2.38E-06

  

0.6

9.74E-46

6.00E-09

6.10E-08

1.92E-08

4.47E-09

 

0.5

0.9

1.28E-10

1.74E-03

3.68E-05

4.56E-03

6.45E-04

  

0.8

4.17E-14

8.92E-06

5.09E-06

1.81E-04

2.94E-05

  

0.7

3.12E-21

4.94E-08

5.46E-10

1.40E-05

2.12E-07

  

0.6

2.19E-21

1.36E-07

2.81E-10

1.20E-06

1.75E-08

 

0.7

0.9

4.20E-06

1.26E-02

1.11E-03

2.86E-02

6.62E-03

  

0.8

1.32E-07

2.00E-04

2.62E-05

3.48E-04

1.15E-04

  

0.7

5.96E-10

2.12E-03

1.49E-04

1.62E-03

4.58E-04

  

0.6

4.06E-10

6.75E-04

3.24E-05

6.96E-04

1.52E-04

Data-M

 

0.9

4.92E-08

8.63E-05

1.97E-05

4.39E-04

4.16E-05

  

0.8

1.01E-09

3.05E-05

5.13E-07

4.84E-05

7.05E-06

  

0.7

2.76E-10

3.31E-05

1.83E-06

1.70E-04

2.28E-05

  

0.6

1.61E-12

8.38E-07

5.29E-09

6.91E-06

6.11E-07

  1. The actual FDR in gene selections across the 100 simulations under each configuration by these methods was compared to that of q-value method using one-tail paired t-statistics.