Table 5

Percentages of significant cases detected (Detected) and false discovery rate (FDR) after multitest adjustment when the p-values come from families of homogeneity tests where some (% effect) of the alternative hypotheses were true.

Strong

SB

BH

SGoF


N

% effect

S

Significant

Detected

FDR

Detected

FDR

Detected

FDR


20

5%

100

7.2

0.26

21

0.29

23

0.38

55

20

5%

1000

7.1

0.04

8

0.05

10

0.77

34

20

5%

10000

7.1

0.00

5

0.02

5

1.67

21

20

10%

100

9.0

0.42

13

0.54

17

1.04

32

20

10%

1000

9.0

0.07

3

0.15

7

2.58

20

20

10%

10000

9.0

0.01

2

0.07

2

3.55

12

20

20%

100

12.5

0.77

7

1.27

8

3.62

17

20

20%

1000

12.7

0.13

2

0.65

3

6.29

10

20

20%

10000

12.8

0.02

1

0.47

2

7.33

7


Mean

0.19

6.9

0.39

8.6

3.03

23.1

SD

0.258

6.51

0.402

7.20

2.45

15.09


40

5%

100

8.4

1.05

7

1.36

12

0.67

26

40

5%

1000

8.5

0.29

2

0.95

6

2.08

14

40

5%

10000

8.5

0.11

1

0.92

2

3.03

6

40

10%

100

11.8

1.98

3

3.12

7

2.90

12

40

10%

1000

11.9

0.58

1

2.76

4

5.50

9

40

10%

10000

11.8

0.21

0

2.83

1

6.42

2

40

20%

100

18.7

3.76

1

8.13

5

9.67

7

40

20%

1000

18.6

1.17

0

7.72

2

12.23

4

40

20%

10000

18.6

0.42

0

7.62

0

13.18

1


Mean

1.06

1.7

3.93

4.3

6.19

9

SD

1.173

2.24

3.029

3.71

4.557

7.70


Strong: The alternative hypothesis is strong.N: sample size. % effect: % of real true alternatives. S: number of tests. Significant: % of significants before adjustment. Detected: % of significant tests after adjustment. FDR: false discovery rate. SB: Sequential Bonferroni. BH: Benjamini and Hochberg. SGoF: Sequential Goodness of Fit.

Carvajal-Rodríguez et al. BMC Bioinformatics 2009 10:209   doi:10.1186/1471-2105-10-209

Open Data