Structural equation modeling of multiple-indicator multimethod-multioccasion data: A primer.

Abstract

[eng] We provide a tutorial on how to analyze multiple-indicator multi-method (MM) longitudinal (multi-occasion, MO) data. Multiple-indicator MM-MO data presents specific challenges due to (1) different types of method effects, (2) longitudinal and cross-method measurement equivalence (ME) testing, (3) the question as to which process characterizes the longitudinal course of the construct under study, and (4) the issue of convergent validity versus method-specificity of different methods such as multiple informants. We present different models for multiple-indicator MM-MO data and discuss a modeling strategy that begins with basic single-method longitudinal confirmatory factor models and ends with more sophisticated MM-MO models. Our proposed strategy allows researchers to identify a well-fitting and possibly parsimonious model through a series of model comparisons. We illustrate our proposed MM-MO modeling strategy based on mother and father reports of inattention in a sample of N= 805 Spanish children.