The outcome variables in this survival analysis was the time (in years) between a conditional starting event and an ending event; this provided a measure of the duration of a given stage. Censoring indicator variables were constructed to identify cases not reaching the ending event by the close of data collection. Three stages in the progression to AD were examined: time to initiation of drinking (age at first full drink), time from initiation to at-risk drinking (years from first drink to either the first intoxication or the first period of regular drinking), and time from at-risk drinking to AD (in years). All age-of-onset variables were obtained by retrospective self-reports. The reliability of retrospective reports of alcohol use has been demonstrated (K. K. Bucholz, Heath, & Madden, 2000; L. B. Koenig, Jacob, & Haber, 2009).
