Since the advent of the genome-wide era, the approach has been plagued by the multiple testing problem, something that was actually recognized before most genome-wide analyses were performed (Zaykin & Zhivotovsky, 2005). Realistic effect sizes may not stand out among a sea of 1 million tests. The problem of separating the genomic wheat from the chaff has led many to question the value of large scale testing. Prevention trials are faced with an additional difficulty in that sampling occurs prior to the onset of disorder or the transition of interest. We performed simulation under a range of effect sizes. While the selection of sample and effect sizes to represent real world settings is important, we present power of the approaches for comparative purposes. That is, while the absolute power of a given sample size at a given effect is interesting, we are interested here only in the relative power of the designs and expect that the relative power of the designs will remain constant across sample size and effect size.
