Multiplicative invalidity and its application to complex correlational models

Area of Science:

• Social Psychology
• Quantitative Psychology

Background:

• Complex correlational models (e.g., path analysis, structural equation models) are increasingly used in social psychology to infer causality.
• Critics highlight issues like model misspecification, inability to infer causality from correlations, and failure to correct for measurement unreliability.

Purpose of the Study:

• To examine the potential for spurious effects in complex correlational models, even under highly favorable assumptions for modelers.
• To assess the impact of measurement validity on the integrity of causal inferences derived from these models.

Main Methods:

• The study employs theoretical analysis, making assumptions maximally favorable to complex correlational modelers by disregarding common criticisms.
• Focuses on how moderately valid measures can introduce artifactual direct and indirect effects.

Main Results:

• Even when common methodological criticisms are disallowed, spurious direct and indirect effects are likely to emerge.
• The validity of measures is shown to be a critical factor in the reliability of complex correlational analyses.

Conclusions:

• Social psychologists should exercise caution when employing complex correlational models.
• Improved handling of measurement validity is essential before widespread adoption of these advanced statistical techniques for causal inference.