Given the large number of statistical tests performed, the likelihood of false positives is high. To address the problem of multiple testing we evaluated the significance level against the overly-stringent Bonferroni threshold (it is questionable whether the Bonferroni correction is appropriate (89), and whether the number of tests are truly independent, given that many SNPs are in high LD). Using the Bonferroni threshold, none of the associations between the 5 personality factors and the individual SNPs tested reached the genome-wide significance. Nevertheless, we believe our results provide useful insights into the genetic architecture of personality traits. For example, when we used simulations that took into account the specific structure of the SardiNIA pedigrees, the availability of phenotype data and the pattern individuals genotyped with the 10K and 500K arrays in each family to evaluate the power of experimental design, we estimated ∼88% power to detect alleles that account for 1.5% or more of the variance in one of the five main NEO-PI-R personality dimensions (at p < 1×10−5). We expect that we would have nearly 100% power to replicate these
