The relationship of scientific objectives to population selection and attrition in longitudinal studies. The case of the normative aging study.