A general approach to representing multifaceted personality constructs: Application to state self‐esteem

TLDR

The article proposes a framework for representing personality constructs at four levels of abstraction. The framework comprises four aggregation models—total aggregation, partial aggregation, partial disaggregation, and total disaggregation—that specify how items and dimensions combine into latent variables. Illustrations of these models on the State Self‑Esteem Scale demonstrate their psychometric properties, invariance, and generalizability.

Abstract

This article proposes a framework for representing personality constructs at four levels of abstraction. The total aggregation model is the composite formed by the sum of scores on all items in a scale. The partial aggregation model treats separate dimensions of a personality construct as indicators of a single latent variable, with each dimension being an aggregation of items. The partial disaggregation model represents each dimension as a separate latent variable, either freely correlated with the other dimensions or loading on one or more than one higher order factor; the measures of the dimensions are multiple indicators formed as aggregates of subsets of items. The total disaggregation model also represents each dimension as a separate latent variable but, unlike the partial disaggregation model, uses each item in the scale as an indicator of its respective factor. Illustrations of the models are provided on the State Self‐Esteem Scale—including tests of psychometric properties, invariance, and generalizability.