Analyses were conducted using SAS 9.2 (SAS Institute, Cary, NC). Kruskal-Wallis χ2 statistics examined mean differences in number of traumatic events and PTSD symptoms by substance dependence status. Logistic regression tested associations between TE load (number of traumatic events) and substance dependence, and PTSD symptoms and substance dependence, controlling for demographic variables (gender, age, immigrant status, and income level). As AD and ND were related to both TE and PTSD, we tested whether relationships between TE load and AD or ND were attenuated after controlling for the effect of PTSD symptoms on AD/ND. We did this by comparing the effect of TE load on substance dependence without PTSD in the model (regression coefficient = τ) to the effect of TE with PTSD in the model (regression coefficient = τ′). A significant difference between τ and τ′ (τ - τ′ > 0) indicates that PTSD symptoms account for at least some of the relationship between TE and AD or ND (41). To generate empirical percentile confidence intervals (CIs) and p values for τ-τ′ since its distribution is unknown (41), we first
