The meta-analytic procedures utilized in the current study followed the recommendations of Lispsey and Wilson (2001) for calculating effect sizes of correlations between two continuous variables. Because there were 350 effect sizes (30 facets and 5 domains × 10 disorders) per matrix, the 38 matrices were first entered into separate spreadsheets where the values were checked by an independent observer to avoid transcription errors. The portion of these matrices concerning each personality disorder were then transposed into separate spreadsheets such that each matrix provided an effect size for the relationship between only that disorder and the 35 FFM variables from all applicable studies. All correlations were then subjected to Fisher's r-to-z transformation. Because many studies contributed more than one correlation matrix (i.e., included multiple measures of DSM and/or FFM constructs) the resulting values were averaged to yield a set of independent effect sizes for each sample. These independent effect sizes were then multiplied by the inverse variance weight of the respective sample (which in the case of Pearson correlation is n-3). These weighted effect sizes were summed for each facet
