Multivariate Behavioral Research · 2019 · 168 citations · 59 references
Social SciencesBehavioral SciencesPsychologyCausal InferencePsychometricsStatistical ModelingFactor AnalysisLatent Trait ModelingLatent Variable ModelingLatent Variable ModelsMeasurement EquivalenceStructural Equation ModelingCausal HypothesesCausal Network ModelingCausal RelationsCausal RelationshipsLatent Variable Causal ModelsObserved VariablesSparse ModelsStatistical Equivalence Criteria
Networks and latent variable models offer competing frameworks for representing psychological constructs, with networks positing direct causal links among observed variables and latent models invoking unobserved common causes; recent work has shown mathematical equivalences between the two, though the constraints they impose differ in interpretability. The authors argue that mathematical equivalence does not guarantee equal plausibility, noting that equivalent constraints may lack clear interpretation across frameworks and that the two approaches yield divergent predictions about the relationship between zero‑order and partial correlations. They propose a likelihood‑based test that compares sparse network models and unidimensional factor models by evaluating their differing implications for the zero‑order versus partial correlation relationship. An empirical example illustrates the test and supports the claim that the two modeling frameworks produce distinct predictions.
Networks are gaining popularity as an alternative to latent variable models for representing psychological constructs. Whereas latent variable approaches introduce unobserved common causes to explain the relations among observed variables, network approaches posit direct causal relations between observed variables. While these approaches lead to radically different understandings of the psychological constructs of interest, recent articles have established mathematical equivalences that hold between network models and latent variable models. We argue that the fact that for any model from one class there is an equivalent model from the other class does not mean that both models are equally plausible accounts of the data-generating mechanism. In many cases the constraints that are meaningful in one framework translate to constraints in the equivalent model that lack a clear interpretation in the other framework. Finally, we discuss three diverging predictions for the relation between zero-order correlations and partial correlations implied by sparse network models and unidimensional factor models. We propose a test procedure that compares the likelihoods of these models in light of these diverging implications. We use an empirical example to illustrate our argument.
59
Construct validity in psychological tests.
Lee J. Cronbach, Paul E. Meehl · Psychological Bulletin · 1955 · 10.9K citations · Full text
Statistical Theories of Mental Test Scores
William W. Rozeboom, Frederic M. Lord, Melvin R. Novick et al. · American Educational Research Journal · 1969 · 8.1K citations