Table 2.

Power of simulation 1

	With length bias						No length bias
	β=0.15			β=0.3			β=0.15
n de=	6	8	10	6	8	10	6	8	10
Weighted	0.36	0.48	0.77	0.81	0.95	0.98	0.61	0.93	0.96
Unweighted	0.29	0.45	0.75	0.77	0.90	0.96	0.61	0.94	0.97
GOSeq	0.23	0.38	0.56	0.97	1.0	1.0	0.96	1.0	1.0
CAMERA	0.01	0	0	0	0	0	0.02	0.01	0.02

Power of identifying set 1 as DE (p-value <0.05) under different β and length bias. The weighted SeqGSA increases the power of the unweighted procedure for detecting set 1. SeqGSA is more powerful with small DE effect (β=0.15), while GOSeq is more powerful with large DE effect (β=0.3). The two procedures of SeqGSA perform similarly under no length bias