Creating short forms for construct measures: The role of exchangeable forms

Abstract

A popular trend has invaded applied psychometrics in terms of creating short forms for construct measures. There seems to be a paucity of developing methodologies for creating short forms based on complete forms that meet psychometric standards related to reliability of the scores and valid inferences. The present article suggests a methodology that rests on the fundamental assumption that the concept of short form attains meaning when being derived from valid scores of a complete form. A pivotal construct for assessing the status of short form is the concept of exchangeable forms which incorporates two types of measurement invariance; invariance across groups frequently exercised in studies applying confirmatory factor analysis, and b) invariance across random facets as estimated in generalizability theory. The two types of measurement invariance involve two types of generalizations relevant for inferring constructs; generalizing from a sample of persons to a population of persons, and generalizing from a sample of construct indicators to a universe or domain of construct indicators. In addition structural invariance is required; exchangeable short forms should relate equivalently to external reference variables. Keywords: Valid complete form; exchangeable forms; generalizing to population of persons and universe of construct indicators; two types of measurement invariance; structural invariance.