TABLE 3

Empirical power of three different association tests in the dominant mode

Θaq+q+/qSample
 size NSignificance
 levelDiplotype
 countingbHaplotype
 inferencecPresent
 algorithm
SAA0.5004.0 2000.0501.0001.0001.000
0.0101.0000.9990.999
0.0010.9960.9920.993
SAA0.5003.0 2000.0500.9990.9990.999
0.0100.9940.9890.991
0.0010.9580.9430.946
SAA0.5002.0 2000.0500.9550.9410.943
0.0100.8550.8230.831
0.0010.6330.5870.600
SAA0.5001.0 2000.0500.0500.0510.051
0.0100.0100.0090.011
0.0010.0010.0010.001
ART0.5004.0 2000.0501.0000.9830.995
0.0100.9980.9380.979
0.0010.9850.8080.903
ART0.5003.0 2000.0500.9950.9310.973
0.0100.9760.8300.909
0.0010.9100.6090.741
ART0.5002.0 2000.0500.9140.7160.810
0.0100.7820.4960.610
0.0010.5310.2460.338
ART0.5001.0 2000.0500.0530.0470.059
0.0100.0090.0100.014
0.0010.0000.0010.001
ART0.2002.010000.0500.9760.8700.923
0.0100.9200.7120.798
0.0010.7860.4660.563
ART0.2001.510000.0500.7450.5460.616
0.0100.5330.3170.382
0.0010.2770.1320.152
ART0.2001.010000.0500.0520.0500.054
0.0100.0100.0100.011
0.0010.0010.0010.001
  • a Θ from SAA or an artificial gene was used. To construct the collection of haplotype copies for the artificial gene, the data were made for six SNPs, each pair of which had weak linkage disequilibrium.

  • b Two-by-two contingency tables were prepared for the χ2 test using the true diplotype configurations of the subjects from simulation data, and the probablity of the tables were calculated by χ2 distribution with 1 d.f.

  • c Two-by-two contingency tables were prepared for χ2 test using the posterior distribution of diplotype configurations estimated by LDSUPPORT, and the probablity of the tables was calculated by χ2 distribution with 1 d.f.

  • Each simulation was performed as described in simulation (under methods) under the alternative hypothesis in the dominant mode with a given penetrance of q+ and varying the other penetrance of q to change the relative risk (q+/q). This simulation was repeated 10,000 times for each parameter set.