Results from simulation study comparing power for rare variant analysis approaches.

<p>500 simulations were based on haplotype distribution for each of 13 deep sequenced candidate genes, and averaged. 500 permutations were run per test. Information for each situation on the bottom of each plot consists of three parts that indicate the test used: (‘C’ for constant, ‘W’ for weighted by allele frequency); (‘’ if signed, ‘’ if constant); and the range of groupings (‘NS’ for nonsynonymous, ‘F’ for all protein coding, ‘F’ for nongenerating protein coding, ‘MAF’ for all MAF, ‘step’ for step-up, and ‘Perf’ for the exact generating alleles when appropriate). Results in plots A-C are sorted by the plot that has the highest area, i.e., the most powerful overall. In D, each value of indicates how much common variants affect disease and must be considered separately; to emphasize this, we have sorted by the power when .