Adjusting for Attrition in School-Based Samples: Bias, Precision, and Cost Trade-Offs of Three Methods

Attrition in longitudinal studies can introduce nonresponse bias when estimating parameters. Methods to correct for nonresponse include survey-based approaches (tracking) as well as analytically based methods (weighting, sample selection modeling). Using data from a multi- wave school-based study of adolescents, substance use estimates are compared across methods. Methods are validated by simulating effects of attrition at baseline, and the relative efficiency of each approach with respect to a known "gold standard" is calculated. Results indicate that weighting may provide sufficient adjustment for nonresponse in other, similar studies. Sample selection modeling requires assumptions that are not met in this setting, and severe bias results. The high costs associated with full tracking efforts may be avoidable, as here we find that tracking was an inefficient approach for bias reduction.