Tracking and attrition in longitudinal school-based smoking prevention research