Multitrait–Multimethod Analysis

Abstract

Multitrait–multimethod (MTMM) analysis is based on the assumption that every measurement in psychology is a trait–method unit, and so contains both trait-related and method-related variance. If each of a set of traits is measured using each of a set of methods, an MTMM matrix of correlations among measures can be calculated. Convergent validation is supported if measures of purportedly the same trait correlate highly across different methods of measurement, discriminant validation is supported if measures of different traits correlate at relatively low levels, and method variance is shown if measures obtained with the same method of measurement correlate highly. Initially, informal rules involving comparisons of correlations within an MTMM were employed to evaluate convergent and discriminant validation. These methods have been supplanted by statistical approaches using confirmatory factor analysis to represent and test patterns of relations among measures. Keywords: data analysis in psychology; psychometric testing; methodology; modeling