Measuring Method Effects: From Traditional to Design-Oriented Approaches

Abstract

This article presents an overview of recent psychometric developments in the area of multimethod measurement, in which we argue that different types of research designs require different types of models. In particular, two types of measurement models for method effects can be distinguished. First, models with a general factor, in which method effects are defined as deviations from a common trait, and second, models for contrasting methods, in which method effects are defined relative to another method but not to a general trait. We argue that the first type of models require a two-level research design (interchangeable methods) whereas the second type of models can be applied to a one-level research design (structurally different methods). Current directions in the uses of these approaches for longitudinal research and multiple-rater studies are described.