Before analyzing the data, we examined them for outliers using an R package (Fox, 2009) that implements the regression diagnostic procedures of Cook and Weisberg (1982). We examined the bivariate distributions of twin scores, and deleted pairs producing significant residuals by Bonferroni-corrected p value. We also deleted pairs producing Cook's distance scores above the threshold suggested by Cook and Weisberg. This resulted in removing 20 pairs (2.1% of the total sample).
