Bayesian Modelling and Algorithms for Heterogenous Interorganisational Networks

No available model combines these sources of heterogeneity into a general framework that admits model uncertainty and allows its probabilistic assessment. The argument that we develop in this proposal is articulated in five steps. First, ERGMs are becoming increasingly popular in the analysis of organisational networks. Second, in this context the empirical applicability of ERGMs is limited by problems of model uncertainty related to specific sources of heterogeneity. Third, a Bayesian approach provides a systematic framework to address core issues of model uncertainty. Fourth, Bayesian models are fought with problems of algorithmic complexity associated with their analytical intractability. Fifth, the project provides solutions to these problems by proposing new methodologies that will support development and application of innovative BERGMs. A crucial part of this project involves the development of specialised software resources. This will prove useful to applied researchers in multiple fields and will be an effective tool for disseminating the results of this research and demonstrate its value. The argument indicates the need for an interdisciplinary approach that combines and integrates substantive interest in ERGMs with: (i) Competencies in Bayesian model development and algorithm design; (ii) Experience in the analysis of empirical data; (iii) Theoretical understanding of interorganisational coordination problems, and (iv) Competencies in the production of software resources. These interdisciplinary sets of competences are well-represented in the research team. The project will be carried out within the SoNAR Centre (sonarcenter.eco.usi.ch), the most recent research unit established at the University of Lugano and relies on the computational facilities of the University of Lugano and Swiss National Supercomputing Centre.