A parsimonious Bayesian predictive model for forecasting new reported cases of West Nile disease  

作　　者：Saman Hosseini Lee W.Cohnstaedt John M.Humphreys Caterina Scoglio 

机构地区：[1]Department of Electrical and Computer Engineering,Kansas State University,Manhattan,KS,USA [2]Foreign Arthropod-Borne Animal Diseases Research Unit National Bio-and Agro-defense Facility,USDA ARS,Manhattan,KS,USA [3]Foreign Animal Disease Research Unit,National Bio-and Agro-defense Facility,USDA ARS,Manhattan,KS,USA

出　　处：《Infectious Disease Modelling》2024年第4期1175-1197,共23页传染病建模（英文）

摘　　要：Upon researching predictive models related toWest Nile virus disease,it is discovered that there are numerous parameters and extensive information in most models,thus contributing to unnecessary complexity.Another challenge frequently encountered is the lead time,which refers to the period for which predictions are made and often is too short.This paper addresses these issues by introducing a parsimonious method based on ICC curves,offering a logistic distribution model derived from the vector-borne SEIR model.Unlike existing models relying on diverse environmental data,our approach exclusively utilizes historical and present infected human cases(number of new cases).With a yearlong lead time,the predictions extend throughout the 12 months,gaining precision as new data emerge.Theoretical conditions are derived to minimize Bayesian loss,enhancing predictive precision.We construct a Bayesian forecasting probability density function using carefully selected prior distributions.Applying these functions,we predict monthspecific infections nationwide,rigorously evaluating accuracy with probabilistic metrics.Additionally,HPD credible intervals at 90%,95%,and 99%levels is performed.Precision assessment is conducted for HPD intervals,measuring the proportion of intervals that does not include actual reported cases for 2020e2022.

关 键 词：West nile virus ICC curve Bayesian model Logistic distribution HPD credible interval 

分 类 号：R511[医药卫生—内科学]