Skip to main content
Fig. 4 | BMC Genetics

Fig. 4

From: Combining controls can improve power in two-stage association studies

Fig. 4

Examples of comparison of power of methods A and B. In all panels, a positive odds ratio corresponds to a deleterious mutation and average MAF is 10%. The top two panels show comparisons of method B with \(n_{0}^{\prime }\) fixed against method A with varying \(n_{0}^{\prime }\). The top left panel has \((n_{0},n_{1},n_{0}^{\prime },n_{1}^{\prime }) = (20169, 5539,8806,6768)\) (values from a GWAS on RA [13]), and the top right panel \((n_{0},n_{1},n_{0}^{\prime },n_{1}^{\prime }) = (15000, 5000,5000,5000)\). Both panels use (α,β,γ)=(5×10−6,5×10−4,5×10−8). The bottom left panel demonstrates the effect of false-ascertainment (F.A) in \(C_{0}^{\prime }\); when cases are mis-ascertained as controls. In this case, (α,β,γ)=(1×10−4,1×10−3,5×10−8), reflecting values used in the paper [14]. The bottom right panel demonstrates a prospective scenario with 10000 samples for replication. Method B with (n0,n1) as above, \((n_{0}^{\prime },n_{1}^{\prime })=(4000,6000)\) is more powerful than any design using method A (grey region; \(n_{0}^{\prime } \in (1000,9000)\); \(n_{1}^{\prime }=10000-n_{0}^{\prime }\))

Back to article page