Timing of initiation was examined as a three-level variable based on participants’ responses to questions about lifetime cannabis use and age at first use: 0=never, 1= onset ≥ age 15, 2= onset < age 15 (asked of everyone at Waves 4 and 5, and those 16+ years old at Waves 1 and 3). Age 15 was chosen to distinguish early- from later-onset users because only 15% of users initiated < age 15, but 15–20% initiated use in each of the subsequent four years (ages 15–18). This definition also satisfied the multivariate normality assumption regarding the distribution of twin and co-twin data (Neale and Maes, 2004) in MZ and DZ twin pairs of both racial/ethnic groups (all p>0.05; range: 0.07–0.84). When reported at multiple interviews, age at initiation was derived from the first interview at which cannabis use was endorsed.
