PANEL ATTRITION AND EXTERNAL VALIDITY IN THE SHORT-TERM FOLLOW-UP-STUDY OF ADOLESCENT SUBSTANCE USE

The present study examines the relationship among grade, substance assessed, and type of measurement and statistical analysis employed in the detection of external validity threats due to panel attrition in the modal follow-up study found in adolescent substance use research. Six successive 2-year panel samples are examined. The results indicate that dropouts demonstrate significantly higher baseline mean use of licit (tobacco and alcohol) and illicit (marijuana and hard drugs) substances when ANOVA is used, but only higher baseline use of illicit substances when chi-square analyses are used. In addition, although mean use tests are no more sensitive to the increased variation in substance use that occurs during adolescence, they are more likely to reveal differences between dropouts and stayers in baseline use of licit substances. The results indicate (a) that findings from school-based, longitudinal studies almost certainly underestimate the level and extent of substance use among adolescents, and (b) that alternative types of attrition analyses yield different effects regardless of grade, even in the short-term follow-up study. These findings are discussed for their implications to adolescent substance use research and for the establishment of standardized procedures to consider the relative risk of Type 2 to Type 1 error in attrition analyses.