ABSTRACT
This chapter is an application of several methods to solve the same problem-the improved understanding of, and correction for, non-random attrition in longitudinal studies. In this chapter, several forms of data analysis are described and used to deal with problems of longitudinal attrition or dropout. Then results from a specific application to real data from the Cognition and Aging in the USA (CogUSA) study are briefly presented. These results illustrate specific features of alternative data analysis models that can be applied to such a problem, including Logistic Regression Models (LRM) and Decision Theory Analysis (DTA). Due to its flexibility and accuracy, the new use of DTA was considered more successful than LRM here, but we know the use of these kinds of correction techniques is not likely to solve the entire problem of longitudinal attrition.
