Permutation-based variance component test in generalized linear mixed model with application to multilocus genetic association study

Table 3 Estimated powers for Simulation 2 with different proportion of the causal markers

n	Method	K	Power with variousm
n	Method	K	0 K	0.1 K	0.3 K	0.5 K
400	Permutation	46	0.052	0.112	0.252	0.402
	Score		0.046	0.103	0.241	0.396
	Mixture (0.50)		0.042	0.096	0.226	0.371
	Mixture (0.65)		0.066	0.130	0.262	0.414
600	Permutation	54	0.053	0.138	0.361	0.583
	Score		0.046	0.125	0.345	0.576
	Mixture (0.50)		0.046	0.121	0.329	0.550
	Mixture (0.65)		0.063	0.153	0.388	0.604
800	Permutation	60	0.049	0.181	0.446	0.707
	Score		0.040	0.173	0.425	0.685
	Mixture (0.50)		0.039	0.149	0.406	0.669
	Mixture (0.65)		0.057	0.192	0.471	0.721
1000	Permutation	65	0.051	0.181	0.528	0.754
	Score		0.046	0.178	0.513	0.746
	Mixture (0.50)		0.040	0.159	0.492	0.727
	Mixture (0.65)		0.057	0.193	0.552	0.773

Note: Permutation is the proposed permutation-based LRT, Score is the score-based sequence kernel association test given in Wu, et al. [31] that was originally developed in Lin (1997), Mixture (0.50) and Mixture (0.65) respectively correspond to the asymptotic 0.50:0.50 and 0.65:035 mixtures of chi-square distributions. Here K is the number of random effects, i.e., the number of SNPs included in a gene, and m is the number of causal SNPs; when m = 0 (i.e., corresponding to 0 K in the fourth column), the estimated power is actually the type I error rate.

ISSN: 1471-2288