Zheng Wei-jun, Li Xiu-yang, Chen Kun
Department of Epidemiology and Biostatistics, College of Medicine, Zhejiang University, Hangzhou 310058, China.
Zhejiang Da Xue Xue Bao Yi Xue Ban. 2008 Nov;37(6):642-7. doi: 10.3785/j.issn.1008-9292.2008.06.017.
Through the multi-stage hierarchical Bayesian model and Markov Chain Monte Carlo methods, Bayesian statistics can be used in dependent spatial data analysis, including disease mapping in small areas, disease clustering, and geographical correlation studies. Recently, Bayesian spatial models have been developed with many types, which have made considerable progress in data analysis. This paper introduces several approaches that have been fully developed and applied, such as BYM model,joint model, semi-parameter model, moving average model and so on. Recently,many studies focused on the comparison work through Deviance Information criterion. Those results show that BYM model and MIX model of semi-parameter model could obtain better results. As more research going on, Bayesian statistics will have more space in applications of spatial epidemiology.
通过多阶段分层贝叶斯模型和马尔可夫链蒙特卡罗方法，贝叶斯统计可用于相关空间数据分析，包括小区域疾病制图、疾病聚集以及地理相关性研究。近年来，已开发出多种类型的贝叶斯空间模型，在数据分析方面取得了显著进展。本文介绍了几种已充分发展并应用的方法，如BYM模型、联合模型、半参数模型、移动平均模型等。最近，许多研究通过离差信息准则进行比较工作。这些结果表明，半参数模型中的BYM模型和MIX模型能获得更好的结果。随着更多研究的开展，贝叶斯统计在空间流行病学应用中将有更大的空间。
