THE ESTIMATION OF RELIABILITY IN LONGITUDINAL MODELS

Despite the increasing attention devoted to the study and analysis of longitudinal data, relatively little consideration has been directed t oward understanding the issues of reliability and measurement error. P erhaps one reason for this neglect has been that traditional methods o f estimation (e.g. generalisability theory) require assumptions that a re often not tenable in longitudinal designs. This paper first examine s applications of generalisability theory to the estimation of measure ment error and reliability in longitudinal research, and notes how fac tors such as missing data, correlated errors, and true score instabili ty prohibit traditional variance component estimation. Next, we discus s how estimation methods using restricted maximum likelihood can accou nt for these factors, thereby providing many advantages over tradition al estimation methods. Finally, we provide a substantive example illus trating these advantages, and include brief discussions of programming and software considerations.