Events

Reciprocity—the tendency for mutual links between nodes—is a common feature in directed networks across social, economic, and scientific domains. In this talk, I present two statistical models that explicitly capture reciprocity while incorporating covariates, with a focus on sparse networks often seen in practice.

The first model is a Bernoulli formulation that enables efficient inference and sheds light on how model components influence estimation accuracy. The second extends the classical $p_1$ framework by allowing both node-level heterogeneity and link-specific reciprocity, supported by a new conditioning-based estimation approach.

Both models are designed for theoretical tractability and practical relevance. I will illustrate their use through examples where reciprocity is central, highlighting their potential applications in business, economics, and beyond.