SMABS 2004 Jena University
SMABS 2004 Home Organization About Jena Sponsors Links Imprint / Contact SMABS Home

European Association of Methodology

Department of methodology and evaluation research

Jena University

Contributions: Abstract

Estimation of maximal reliability via covariance structure modeling: Recent extensions

Tenko Raykov
Research Institute on Addictions, Buffalo, NY

This talk initially outlines a covariance structure modeling approach to the problem of estimating maximal reliabilty for a weighted composite of a given set of congeneric measures with uncorrelated errors. The method represents a single-session modeling procedure that also simultaneously yields standard errors for the weights in the optimal linear combination of the measures. As an extension, a readily and widely applicable approach to interval estimation of the maximal reliability coefficient is outlined that is based on a first-order Taylor expansion of the latter. The relationship of maximal reliability to a factor score estimator, its variance, and a factor determinacy coefficient is then discussed. Confidence intervals for individual factor scores are finally considered that depend on the maximal reliability coefficient.

Keywords: congeneric measures, covariance structure modeling, maximal reliability