Classification With Longitudinal Data

• In Event: Advances in Longitudinal Modeling

Abstract

Classification, the act of placing examinees into given categories based on a psychometric measure, was explored in the context of longitudinal data. The fitted values of a properly specified multilevel model led to more accurate classifications than classifications based on the final wave of observed data. To represent the uncertainly in the fitted values, a simulation procedure gave the standard error of the fitted values and was then used to successfully extend expected classification accuracy (Rudner, 2001, 2005) to the longitudinal situation. Further, there was support to make classifications about future time points and obtain an estimate of the accuracy of these predictive classifications.

Author

• Quinn Nathaniel Lathrop, University of Notre Dame