Latent Interaction Modeling with Ordinal Items: Evaluating Alternative Analytic Methods and Parceling Strategies

The present study investigated two popular latent interaction modeling methods, unconstrained product indicator (UPI) and latent moderated structural equations (LMS) when parceling ordinal items. Two parceling strategies were considered: (1) combining items with similar loadings into the same parcel (Isolated strategy) and (2) distributing items with similar loadings across parcels (Balanced strategy). Additionally, we employed an Aggregated strategy that considers item-to-parcel-allocation variability by aggregating results across different allocations. A Monte Carlo study compared UPI and LMS under different parceling strategies. The manipulated factors were the interaction effect size, factor loadings, number of response categories, item distributions, and sample size. All approaches demonstrated satisfactory performance in estimating latent interaction effects when items had four or more response categories. We recommend UPI with the Balanced and the Aggregated strategies because they generally showed comparable or better performance in interaction effect estimation compared to other approaches. Suggestions for future research are discussed.