Deriving Estimators and Their Standard Errors in Dyadic Data Analysis: Examples Using a Symbolic Computation Program
Michael Browne’s foundational papers on the analysis of covariance structures (e.g., Browne, 1974, 1982) were inﬂuential in our conceptualization of models for dyadic data analysis. He provided a framework for working with covariance structures that makes it relatively easy to derive new results. We began working on dyadic models in the early 1990s, before the explosion of research on multilevel models and random effect models. Motivated by the ubiquity of dyadic data in social psychological research, we recognized that a structural equation modeling (SEM) approach with latent variables representing dyadic similarity would be helpful in modelingmultivariate dyadic data.Multilevelmodels, let alonemultivariatemultilevel models, were relatively new at that time and we explored their usefulness for dyadic data analysis. But it was the SEM approach and the Browne (1974, 1982) papers that we found most helpful in our initial model formulation.