Two-mode (bipartite) networks can be used to represent affiliations of individual actors with organizations, meeting places, preferences, etc., while the more usual one-mode networks can represent affective, cooperation, exchange, etc. relations between these actors.
If for a given group of actors relevant longitudinal data are available about (e.g.) a one-mode cooperation network and a two-mode affiliation network, it is plausible that affiliation patterns will influence cooperation relations, and cooperation relations will influence affiliation patterns. A continuous-time Markov chain model is proposed for the simultaneous dynamics of a one-mode and a two-mode network, and some specifications of the structural influences are discussed. Estimation procedures (method of moments, maximum likelihood, Bayesian) are briefly discussed. This is applied to panel data of friendship and advice relations (two one-mode networks) in a group of MBA students and their preferences for future employers (one two-mode network).
(Work in collaboration with Alessandro Lomi and Vanina Torlo)