In an effort to reduce potential retrospective reporting bias (i.e., the trend toward reports of age at first use to increase slightly with respondent age [Grucza et al., 2008]), adjustments were made for the length of time between age at first use and age at the time of report. Since an individual’s reports of first alcohol, cigarette, and cannabis use could be derived from different waves of data collection, separate adjustments for time between first use and report of first use were made for each substance. A total of 10 dummy variables per substance, representing 0, 1, 2, up through 10 or more years of lag time, were included in analyses. (Dummy variables were deemed a better fit than continuous variables, which rely heavily on assumptions of linearity that are not met in our measures of lag times.) Lag times for those women who never reported any use of a given substance were coded as 0 for that substance since assignment to the ‘never-used’ category was based on reports from the most recent interview.
