Table 1.

Performance of the Subset-Based Test for Detection of the Truly Associated Subset of Traits

(K, T1, T2)	Sensitivity (True Positive Probability)		Specificity (True Negative Probability)
	One-Sided	Two-Sided	One-Sided	Two-Sided
5 traits: 100% positive
(5, 1, 0)	0.920	0.986	0.835	0.500
(5, 2, 0)	0.943	0.943	0.886	0.505
(5, 3, 0)	0.934	0.935	0.921	0.477
(5, 4, 0)	0.924	0.925	0.926	0.412
5 traits: 75% positive
(5, 1, 1)	0.502	0.982	0.883	0.721
(5, 2, 1)	0.621	0.971	0.906	0.745
(5, 3, 1)	0.676	0.950	0.918	0.782
10 traits: 100% positive
(10, 2, 0)	0.942	0.953	0.879	0.563
(10, 3, 0)	0.941	0.944	0.893	0.579
(10, 4, 0)	0.932	0.932	0.914	0.587
(10, 5, 0)	0.926	0.927	0.923	0.573
(10, 6, 0)	0.927	0.927	0.922	0.557
(10, 7, 0)	0.922	0.924	0.933	0.522
10 traits: 75% positive
(10, 1, 1)	0.500	0.978	0.855	0.675
(10, 2, 1)	0.621	0.973	0.877	0.698
(10, 3, 1)	0.672	0.955	0.901	0.739
(10, 3, 2)	0.566	0.934	0.907	0.778
(10, 4, 2)	0.616	0.924	0.915	0.801
(10, 5, 2)	0.656	0.933	0.927	0.825

In each simulation, a total of K = 5 or K = 10 distinct traits are analyzed (each trait has 2,000 cases and 2,000 controls). A variant of MAF = 0.3 is assumed to be associated with a subset of size T (<K) of the traits with an odds ratio of 1.15 (see Table S1 for results under heterogeneity of odds ratios). The “100% positive” sections assume that all of the associations are in the same direction, and the “75% positive” sections assume that 75% of the associations are positive and 25% are negative. The two measures of performance that are shown are (1) sensitivity (the average proportion of associated traits detected) and (2) specificity (the average proportion of null traits discarded). The following abbreviations are used: K, total number of traits analyzed (5 or 10); T₁, number of traits that are truly associated in the positive direction; and T₂, number of traits that are truly associated in the negative direction.