Minimum Redundancy Item Analysis

Publisher Website
Google Scholar
Abstract
The problem of constructing tests showing high levels of reliability and minimum mutual redundancy is considered in terms of item analysis and optimization. Certain item indices designed to optimize homogeneity while reducing redundancy are evaluated and compared. An analysis was undertaken with the aim of reducing 40-item to 20-item scales with optimal levels of the desired properties of mutual independence and homogeneity. Twenty sets of parameters yielded different levels of redundancy, evaluated in terms of the size of the first principal component, the number of 'large' components, and Gini's coefficient of concentration. It was found that item selection procedures can markedly reduce inter-scale redundancy while maintaining optimal levels of homogeneity.