Bayesian Spatial Predictive Synthesis
Bayesian Analysis · 01.01.2026 · -1 (-1) · 1-27
DOI: 10.1214/26-ba1625 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bayesian Analysis, erschienen am 01.01.2026. Verfasst von Danielle Cabel und 4 weiteren Autor:innen. Die Arbeit nennt 37 Literaturangaben.
Abstract (Originalsprache)
We propose a Bayesian ensemble methodology to synthesize multiple predictive distributions of spatial data. Ensemble methods are a common approach to mitigate model specification in statistical analysis. Due to spatial dependence, typical ensemble models, which assume i.i.d. data, can cause bias when there is data and model heterogeneity. Our methodology is defined by specifying a latent factor spatially-varying coefficient model as the synthesis function within the Bayesian predictive synthesis framework, which enables spatial characteristics of each model to be learned and ensemble coefficients to vary over regions to achieve flexible predictions. Theoretical developments underpin our approach, where we show our method is the best approximation of the data generating process. Posterior inference is conducted via Markov chain Monte Carlo, with both standard and scalable implementations for full uncertainty quantification, along with a variational inference approach for fast point estimation. Extensions to general responses are also developed. Simulation studies and two real data applications in real estate and ecology highlight the strengths of our method, outperforming standard spatial models and ensemble methods, and advanced machine learning methods, in terms of predictive accuracy and uncertainty quantification.
Zitieren
APA
Cabel, D., Sugasawa, S., Kato, M., Takanashi, K., & McAlinn, K. (2026). Bayesian Spatial Predictive Synthesis. Bayesian Analysis, -1(-1), 1-27. https://doi.org/10.1214/26-ba1625
Vancouver
Cabel D, Sugasawa S, Kato M, Takanashi K, McAlinn K. Bayesian Spatial Predictive Synthesis. Bayesian Analysis. 2026;-1(-1):1-27. doi:10.1214/26-ba1625
BibTeX
@article{Cabel2026,
author = {Cabel, Danielle and Sugasawa, Shonosuke and Kato, Masahiro and Takanashi, Kōsaku and McAlinn, Kenichiro},
title = {Bayesian Spatial Predictive Synthesis},
journal = {Bayesian Analysis},
year = {2026},
volume = {-1},
number = {-1},
pages = {1--27},
doi = {10.1214/26-ba1625}
}
EUVIMED