of the sample, power must lie between that for a sample of 2300 (i.e., the number of independent families) and 7188 (i.e., the number of individuals) observations. Consequently, with a two-tailed test at p < 5 × 10-8, we have power of at least 80% to detect effects with r2 between .005 and .017. Given GWAS findings with other quantitative traits (Visscher et al., 2012), the lower bound of this interval is on the high end of what might be expected. In any case, although we failed to identify variants with p values below the standard GWAS threshold, Panagiotou et al. (2012) have recently shown that a high proportion of associations with borderline significance in an initial analysis (which they defined as a p value < 1 × 10-7) go on to be confirmed in subsequent GWAS. The 13 variants given in Table 2 that showed effects that were consistent across the five phenotypes represent candidates we believe worthy of future consideration.
