A Dyadic IRT Model

Area of Science:

• Psychometrics
• Social Psychology
• Statistical Modeling

Background:

• Traditional Item Response Theory (IRT) models individual responses.
• Social Relations Model (SRM) analyzes dyadic interactions but lacks IRT's measurement properties.
• Existing models often fail to capture the nuanced, directional nature of dyadic relationships.

Purpose of the Study:

• To propose a novel dyadic Item Response Theory (dIRT) model for measuring interactions within pairs.
• To generalize existing IRT and SRM frameworks for dyadic data analysis.
• To provide a flexible framework for analyzing complex relational data.

Main Methods:

• Developed a dIRT model incorporating actor, partner, and unique dyadic effects.
• Utilized Markov-chain Monte Carlo (MCMC) estimation via Stan for model implementation.
• Extended the dIRT model to include latent regressions and multilevel structures.

Main Results:

• The dIRT model successfully captures actor tendencies, partner elicitations, and unique dyadic relationship variables.
• Estimation procedures demonstrated good performance in simulation studies.
• Analysis of speed-dating data revealed significant pairwise interaction effects on mutual attraction.

Conclusions:

• The dIRT model offers a powerful and flexible approach to understanding dyadic interactions.
• Pairwise interactions are crucial and cannot be solely explained by individual characteristics.
• The proposed model has broad applications in social sciences and behavioral research.