Font Size:
Marginal modeling of multilateral relational events
Last modified: 2017-05-20
Abstract
We implement the methodology of marginal modeling of relational events involving groups of actors, as developed in Bartolucci, Peluso, Mira (2017). Current relational data analyses suffer from the representation of an event through edge variables, with potential loss of information when the events generate a set of multiple relations rather than bilateral connections or ties. To fully exploit the informational content of relational events, we model an event as a binary vector of response variables representing actors partecipating to the event. Univariate and bivariate distributions of the events are modeled through marginal parameters having a clear social interpretation.