Table 2. Power of META-GSA, pooledGWAS-GSA and SPP across all studies, based on 100 genes with 1 marker each.

patterns of true marker-phenotype associations
scenario no.		RRs in GS of interest	RRs in complementary GS´	no. studies	SPP	pooledGWAS-GSA	META-GSA
H 0 : F(RR|GS) = F(RR|GS´)
1	no gene is associated at all	10x1	90x1	10	5.6%	3.7%	4.4%
2	all genes are associated	10x1.2	90x1.2	10	5.2%	5.5%	7.0%
3	all genes are associated	10x1.5	90x1.5	10	6.0%	5.6%	4.8%
H A : RR>1 only in GS / F(RR|GS)≠F(RR|GS´)
4	½ the genes in GS are associated	5x1 5x1.1	90x1	10	5.4%	6.7%	9.2%
5		5x1 5x1.2	90x1	10	9.4%	9.5%	26.4%
6		5x1 5x1.3	90x1	10	21.0%	20.7%	48.8%
7		5x1 5x1.4	90x1	10	29.6%	31.3%	58.0%
8		5x1 5x1.5	90x1	10	36.8%	56.4%	58.8%
H A : RR>1 only in GS / F(RR|S)≠F(RR|NS)–increasing number of studies
9		5x1 5x1.5	90x1	2	16.0%	12.7%	24.4%
10		5x1 5x1.5	90x1	3	16.8%	15.9%	31.2%
11		5x1 5x1.5	90x1	4	27.2%	21.3%	41.8%
12		5x1 5x1.5	90x1	5	27.8%	25.0%	45.0%
13		5x1 5x1.5	90x1	6	28.4%	30.2%	47.4%
14		5x1 5x1.5	90x1	7	31.8%	38.2%	50.4%
15		5x1 5x1.5	90x1	8	34.4%	40.9%	54.4%
16		5x1 5x1.5	90x1	9	35.6%	47.2%	55.2%
H A /H 0 : mixed structured GS
17	HA: assoc. genes in GS only	5x1 4x1.2 1x1.5	90x1	10	17.0%	16.1%	35.6%
18	HA: GS dominates GS’	5x1 4x1.2 1x1.5	63x1 18x1.2 9x1.5	10	11.0%	11.4%	5.8%
19	HA: GS is dominated § by GS‘	5x1 4x1.2 1x1.5	41x1 32x1.2 17x1.5	10	8.4%	1.7%	1.6%
20	H0: same prop. of genes are associated in GS and GS’	5x1 4x1.2 1x1.5	45x1 36x1.2 9x1.5	10	4.4%	4.7%	3.0%

Given a true type I error of 5%, the observed type I error may range from 3% to 7% (95% random dispersion interval for 500 simulations). Given a true power of 50%, the observed power may range from 45% to 54% (95% random dispersion interval for 500 simulations).

^§ Truly associated genes are more frequent in GS’ than in GS.