TABLE 2

Effect of sample size on LD metrics and their significance

Jackknifed subsamplesaObserved
N = 64, ± (SD)N = 128, ± (SD)N = 255, ± (SD)ABC
Tests5797.4 (1949.4)5649.1 (622.7)5662.5 (29.4)5671.05460.013530
Average r2   0.068 (0.006)   0.062 (0.004)   0.058 (0)   0.058   0.056    0.040
Average D   0.388 (0.096)   0.394 (0.039)   0.400 (0.001)   0.400   0.252    0.464
>0.0576.4% (0.026%)62.2% (0.024%)47.3% (0.002%)47.3%61.4%63.2%
<0.0523.6% (0.026%)37.8% (0.024%)52.7% (0.002%)52.7%38.6%36.8%
<0.0113.3% (0.017%)24.5% (0.019%)39.2% (0.002%)39.4%27.7%24.8%
<0.0017.4% (0.01%)15.3% (0.014%)27.8% (0.002%)27.9%21.0%16.3%
<0.0001 4.9% (0.009%)10.7% (0.009%)20.5% (0.001%)20.6%16.8%11.8%
<0.00001 3.5% (0.008%) 8.2% (0.007%)16.1% (0.001%)16.2%14.1% 9.0%
<0.000001* 2.4% (0.006%) 6.6% (0.005%)13.0% (0.001%)13.1%12.4% 7.0%
<0.0000001 1.7% (0.005%) 5.4% (0.005%)10.5% (0.001%)10.5%10.8% 5.5%
  • * The Bonferroni significance threshold is 0.000001 for 5000 tests.

  • a Based on 25 random subsamples from the full allele matrix. Sites were extracted on the basis of the minimum number of polymorphic alleles for a site and the absolute count of the less common allele. For N = 64, only sites represented by 50 or more alleles were included, if and only if the rare variant was present in 5 or more alleles. These numbers were 100 and 10 for N = 128, or 200 and 20 for N = 255. The observed sets were: A, 200, 20; B, 150, 20; and C, 200, 5.