Divorce was modeled using a single threshold, identified by fixing the mean to 0, variance to 1, and estimating the threshold. All other variables were treated as continuous and standardized to mean=0, SD=1. Indirect effects were calculated for all paths from parental polygenic scores to offspring alcohol use outcome through mediator variables, for a total of 8 indirect effects per model. Confidence intervals for indirect effects were calculated via bootstrapping with 1000 replications.
