Congratulations to Xiaotian Zheng from the Centre for Environmental Informatics at the University of Wollongong! His paper "Bayesian geostatistical modeling for discrete-valued processes", co-authored by Athanasios Kottas, Bruno Sansó has been awarded the 2023 TIES best student paper award.
His work introduces a novel nearest-neighbour mixture process (NNMP) in which data comes from a multivariate discrete distribution that can be represented by a directed acyclic graph over a continuous spatial domain. Instead of modelling spatial correlations through the traditional class of spatial generalized linear mixed models, data reduction is performed via a mixture model in which the mixing probabilities arise from a copula model, and the spatial correlations are embedded in the associated copula. The model hinges on two propositions that allow the full joint distribution to be factored into the product of a series of spatial marginal distributions (from the copula) and the conditional (joint) distribution given the spatial locations. Inference is challenging and an MCMC algorithm that uses Metropolis updates for some of the parameters is developed. In summary, the proposed discrete NNMP provides a novel process-based modelling approach for analyzing spatially correlated discrete data in a computationally efficient way. The methodology was used in an interesting ecological application using real world data from a bird population monitoring survey.
His paper was well written with large potential for applications in the analysis of spatially correlated discrete data. The proposed methods have a potential for broader use and provide a significant contribution to research in the field of Environmetrics.
