In order to obtain a subsample of unrelated individuals, we filtered the sample using genetic relatedness thresholds of .025, .05, and .1 (Yang et al., 2013), the most stringent of which corresponds approximately to the relationship between third or fourth cousins. Results for each subsample are given in Table 3, which also presents results for the same subsamples using a GRM based on SNPs weighted by LD patterns (Speed et al., 2012). The standard errors are large, a consequence of the fact that unrelated individuals are used, which also reduced the sample approximately in half. SNP heritability estimates vary somewhat, as would be expected due to sampling error across the subsamples, but are relatively consistent. The median estimates were .29 and .27 for P3 amplitude and the genetic factor score derived from parameters of the biometric common factor model, respectively. Standard errors in some cases were larger than the point estimates, making strong inferences ill advised. Table 3 also includes SNP heritability estimates for each endophenotype in the full sample (i.e., without imposing a threshold of relatedness, appearing in the
