Introduction

Abstract

The importance of Generalizability theory (G theory) lies in its applications to educational measurement. Two of its major functions are: 1) to evaluate the quality of measurement procedures; and 2) to make projections about how one can improve the quality of measurement procedures. Regardless of its wide applications (Brennan, 1997, 1998; Lane, Ankenmann, & Stone, 1996; and Linn, Burton, DeStefano, & Hanson, 1996), G theory, a framework relying on the estimation of variance components, has a major limitation in its capability of handling missing data — a common problem in large-scale assessments. Test developers often cannot use ordinary algorithms for estimating variance components in G theory because of the excessive computational requirements.