The correlation between Time 1 and Time 2 age at first drink reports was calculated using the raw (continuous) data after removing extreme outliers (the highest and lowest 1% of the range of Time 2-Time 1 difference scores). As a second approach to assessing consistency in reports, specifically, consistency in categorization of drinkers as early vs. non-early initiators, a dichotomous variable representing early use was created, with the lowest one-third of the distribution of age at first drink in Time 1 reports (14 years or younger) coded positive and the remainder (‘average to late’ first users) as negative for early use. A tetrachoric correlation coefficient was calculated to determine consistency of early initiation status across time points (using the same definitions at Time 1 and Time 2).
