An empirical evaluation of alternative approaches to adjusting for attrition when analyzing longitudinal survey data on young adults' substance use trajectories