The proportional hazards assumption was satisfied in first DSM-5 symptom models. Violations were noted in a subset of first intoxication analyses. Examination of Scholenfeld residuals indicated that the group of 17 year olds was driving this violation, perhaps reflecting important transitions at this age. Censoring at age 17 instead of 18 satisfied the proportional hazards assumption without substantially altering the parameter estimates, supporting our conclusions.
