Date: 2023-04-27

Time: 14:00-15:00 (UK time)

Strand S2.29

Abstract

We focus on models for network data which contain quantitative information about relationships between nodes. In particular, we consider model that capture similarity between nodes through unobserved latent variables. We investigate how to deal with model specification uncertainty through formal Bayesian model averaging. In the context of a Poisson model with a log-Normal rate parameter, we give conditions under which a common improper prior leads to posterior existence. We derive an MCMC strategy for inference using Bayesian model averaging over the model space (constructed by including or excluding each of the covariates, including the latent ones). The model is applied to bilateral migration flows between 38 OECD countries during the period 2015-2020 and it is shown to outperform three popular gravity models.

This is joint work with Gregor Zens.

Speaker

Mark Steel is a professor at the Department of Statistics, University of Warwick. He is interested in theoretical and applied Bayesian statistics, particularly distribution theory, Bayesian model averaging, spatial statistics, non- and semiparametric inference, survival models, stochastic frontier models, contingent valuation and stochastic volatility models. He is currently the Editor-in-Chief of Bayesian Analysis.