We examined potential correlates of use of any substance among youth (i.e., collapsing across substances) during the pandemic. For each correlate, we fit a logistic regression to all available observations across the three waves using the survey package [32] in R, clustering on site, family, and youth to account for nonindependence of observations. Separate regressions were fit for each substance. Youth age and dummy variables for survey wave were included as covariates. For each correlate, we tested for moderation [33] of the association with youth use of any substance by youth sex and age. Given the multiple tests, we applied the false discovery rate (FDR) adjustment [34] and report adjusted p values in all instances in which a given association was no longer statistically significant at p < .05.
