A0158
Title: Model uncertainty in latent Gaussian models with univariate link function
Authors: Mark Steel - University of Warwick (United Kingdom) [presenting]
Gregor Zens - IIASA (Austria)
Abstract: A class of latent Gaussian models are considered with a univariate link function (ULLGMs). These are based on standard likelihood specifications (such as Poisson, Binomial, Bernoulli, Erlang, etc.) but incorporate a latent normal linear regression framework on a transformation of a key scalar parameter. Model uncertainty regarding the covariates included in the regression is allowed. The ULLGM class typically accommodates extra dispersion in the data and has clear advantages for deriving theoretical properties and designing computational procedures. Posterior existence is formally characterized under a convenient and popular improper prior, and an efficient Markov chain Monte Carlo algorithm is proposed for Bayesian model averaging in ULLGMs. Simulation results suggest that the framework provides accurate results that are robust to some degree of misspecification. The methodology is successfully applied to measles vaccination coverage data from Ethiopia and to data on bilateral migration flows between OECD countries.
