Estimating reliability coefficient for multidimensional measures: A pedagogical illustration

Abstract

illustration WAHYU WIDHIARSO and HAMDOLLAH RAVANDThe literature has shown that the assumption of unidimensional measurement in psychology is difficult to fulfill, since measurement in psychology is usually a complex process that aims at providing information about constructs.As a consequence, factor analysis for psychological measurement tends to conceal several factors that underlie the items on the scale.Since applying a reliability coefficient (e.g., Cronbach's alpha) based on a unidimensional assumption for a multidimensional measure will underestimate reliability, researchers should use an appropriate coefficient that matches the characteristics of the measure.There are several, albeit not frequently utilized reliability coefficients for multidimensional measures.The present article demonstrates the application of the stratified alpha, Mosier's, Raykov's, McDonald's, and Hancock-Mueller's coefficients for estimating the reliability of multidimensional measures.