Prevalences and standard errors were estimated using SUDAAN to adjust for the complex sample design. 39 Weighted correlations were produced in SAS. Weighted exploratory factor analysis (EFA) was conducted using Mplus 40 using Promax rotation, allowing correlated factors. EFA results were used to construct observed cannabis withdrawal symptom count variables. To examine how indicators of clinical significance and aspects of predictors related to cannabis withdrawal as dependent variables, weighted negative binomial regression models were conducted with STATA. 41 Demographic variables were included as needed based on association with clinical significance indicators and outcome variables. The regression coefficient for a predictor is the log ratio of the change in means in outcome for a unit change in the predictor. Therefore, the coefficients were exponentiated to provide ratio estimates interpreted similarly to an odds ratio.
