Center Director, Research Center for Group Dynamics, Institute for Social Research
Director, BioSocial Methods Collaborative, RCGD
Amos N Tversky Collegiate Professor, Psychology and Statistics, LSA
Professor of Marketing, Stephen M Ross School of Business
Professor of Integrative Systems and Design, College of Engineering
Analyzing multivariate dyadic data for exchangeable dyads
This was our first paper on dyadic analysis. In retrospect, one of our main contributions was more pedagogical in that we showed how to get solid intuition about dyadic data and presented a framework in which to derive estimates and their standard errors. The framework illuminated several aspects of dyadic data, including why a correlation of dyadic means is sometimes difficult to interpret. Our results are identical to multilevel models using maximum likelihood estimation.
Griffin, D. & Gonzalez, R. (1995). Correlation models for dyad-level models: I. Models for the exchangeable case. Psychological Bulletin, 118, 430-439. \doi: 10.1037/0033-2909.118.3.430
Many research problems in psychology require statistical methods that take into account the dependencies introduced by dyadic interaction. The authors provide correlational tools for dyadic data when the individuals within the dyads are both from the same class or category, such as 2 male adults. First, the authors provide significance tests for correlations between 2 variables when individuals are nested within dyads. Second, they provide a simplified method for decomposing the overall correlation into individual-level and dyad-level relations. Finally, the authors demonstrate these methods with dyadic data collected by L. Stinson and W. Ickes (1992) in a study of unstructured dyadic interactions.