Interazioni, regressioni, LISREL: Sviluppi nell’analisi confermativa delle relazioni moltiplicative

We analyzed different parametrizations of structural equation models including interaction terms. The models showed herein (e.g., Joreskog, 1998) are less complex and less computing-demanding than previously proposed parametrizations (e.g., Jaccard & Wan, 1995). Moreover, they have some advantages over the classical multiple regression models using observed variables. We used data from a study on dieting intentions (N = 609 adult women). The theoretical model hypothesizes that intentions to stick to a diet are a function of the main effects of Subjective Norms, Self-efficacies concerning the ability to resist eating temptations, and of the interaction between subjective norms and self-efficacy. Results show that structural equation models outperform hierarchical regression models estimating the interaction effect with higher precision. Pros and cons of the models tested, and avenues for a wider implementation of such models in substantive psychological research are discussed.