At 11-, 14-, and 17-year-old time-points, a computerized substance use questionnaire was individually self-administered, without the presence of project staff. Questions were phrased as, “Have you ever used [substance]?” Estimation of IRT parameters requires that at least one person per group endorses each item (in this case, groups were defined by age, sex, and twin birth order). To achieve this in each group and to maintain conformity across substances, responses to use of each substance were converted to dichotomous yes/no values (original response options differed between substances) and substances with lower endorsement rates were clustered into substance classes. The final measure analyzed here was which substances (tobacco, alcohol, and marijuana) or classes of substances [other controlled substances (stimulants, tranquilizers, Quaaludes/downers, cocaine, PCP/LSD/other psychedelics, and opiates) and uncontrolled substances (over-the-counter medications and inhalants)] the participant endorsed having ever used (at the intake assessment) or used since the previous assessment (at each follow-up) recreationally. This approach is useful in this context as it retains variance at all ages (which is especially challenging when dealing with the relatively low usage rates among 11-year-olds).
