Table 3 Performance of SAMC on extreme p-values

From: Detecting heritable phenotypes without a model using fast permutation testing for heritability and set-tests

Site

Parametric p-value

Perm. p-value

CI

SAMC, N = 105

SAMC, N = 106

cg00123214

2.601  10−08

12  10−09

(6.201, 20.962)  10−09

6.301  10−09

13.71  10−09

cg00044796

3.779  10−13

4  10−09

(1.090, 10.242)  10−09

6.092  10−09

5.065  10−09

cg01821635

3.941  10−18

2  10−09

(0.242, 7.225)  10−09

4.500  10−09

6.504  10−09

cg06784218

5.756  10−21

4  10−09

(1.090, 10.242)  10−09

2.327  10−09

6.484  10−09

cg08963013

4.625  10−26

6  10−09

(2.202, 13.059)  10−09

3.339  10−09

5.460  10−09

cg14252149

5.005  10−34

4  10−09

(1.090, 10.242)  10−09

3.358  10−09

5.131  10−09

cg00071950

5.678  10−36

5  10−09

(1.623, 11.668)  10−09

8.149  10−09

7.440  10−09

cg02002194

9.238  10−42

4  10−09

(1.090, 10.242)  10−09

6.420  10−09

7.501  10−09

cg11266682

1.873  10−46

7  10−09

(2.814, 14.423)  10−09

12.62  10−09

6.251  10−09

cg15296535

8.427  10−51

9  10−09

(4.115, 17.085)  10−09

6.839  10−09

9.045  10−09

  1. For 10 selected sites, we show the parametric p-value, as calculated by GCTA; the permutation p-value, as calculated by 109 permutations, and the CIs implied by it; and the p-values estimated by SAMC, using N = 105 or N = 106 permutations (with t0 = 1000). SAMC estimates, even for N = 105, appear accurate for all sites, as they lie within the CI calculated from the permutation test