Empirically Supported Methodological Critique of Double Entry in Dyadic Data Analysis

Analyzing dyadic phenomena (e.g. trust, power, and satisfaction) gains importance not only in sociology and psychology, but also in economics and management. The aim of the paper is to examine the mathematical foundation of Dyadic Data Analysis (DDA). On one hand, we critique the database development of DDA for exchangeable cases, and develop an algorithm for transforming such a data set into distinguishable cases. On the other hand, we question the usefulness of a widely used data development technique of DDA, the so-called double entry. We reason that this technique does not necessarily lead to additional information. In contrast, it might lead to information losses. We develop approximations for correlations and regression models of DDA. These are also empirically tested using a database of 89 dyads. The obtained results back our theoretical reasoning, most of the approximations give satisfying results. This support our main proposition that mathematical foundation of DDA needs further research.