0 and 1 (Nakagawa & Schielzeth, 2012); in COGA, the ‘MuMIn’ package in R was used to calculate marginal R2 for the logistic mixed-effect models for the binary outcomes (Barton, 2011). Once the most predictive AUDIT-C and AUDIT-P PRS thresholds were determined in the single-PRS models (n.b. pT was allowed to vary between AUDIT-C and AUDIT-P PRS), the most predictive PRS for both AUDIT-C and AUDIT-P PRS were simultaneously entered into a joint regression model along with covariates (results for single-PRS models are in available in Supplemental Table 1–12). We primarily report on the results of the joint (AUDIT-C PRS + AUDIT-P PRS) regression models, as we wished to examine the relative contribution of each AUDIT subscale PRS while controlling for the other subscale PRS. As the primary analyses consisted of 2 tests (AUDIT-C PRS and AUDIT-P PRS) for each of the 12 outcomes, and the number of independent tests across the PRS p-value thresholds was estimated to be approximately 5 (calculated using spectral decomposition, via the matSpD.R R script (Nyholt, 2004)), we corrected for 120 tests using a Bonferroni p-value = 0.0004.
