Department of Statistics, University of Akron, Akron, OH, USA.
Department of Environmental Health Science, Indiana University-Purdue University, Indianapolis, IN, USA.
Spat Spatiotemporal Epidemiol. 2020 Aug;34:100360. doi: 10.1016/j.sste.2020.100360. Epub 2020 Jul 16.
In this paper, we compare a variety of spatio-temporal conditional autoregressive models to a dengue fever dataset in Colombia, and incorporate an innovative data transformation method in the data analysis. In order to gain a better understanding on the effects of different niche variables in the epidemiological process, we explore Poisson-lognormal and binomial models with different Bayesian spatio-temporal modeling methods in this paper. Our results show that the selected model can well capture the variations of the data. The population density, elevation, daytime and night land surface temperatures are among the contributory variables to identify potential dengue outbreak regions; precipitation and vegetation variables are not significant in the selected spatio-temporal mixed effects model. The generated dengue fever probability maps from the model show a geographic distribution of risk that apparently coincides with the elevation gradient. The results in the paper provide the most benefits for future work in dengue studies.
在本文中，我们将多种时空条件自回归模型与哥伦比亚登革热数据集进行了比较，并在数据分析中采用了一种创新的数据转换方法。为了更好地了解不同生态位变量在流行病学过程中的影响，我们探索了泊松-对数正态和二项式模型，并采用了不同的贝叶斯时空建模方法。我们的结果表明，所选模型能够很好地捕捉数据的变化。人口密度、海拔、白天和夜间陆面温度是确定潜在登革热爆发区域的贡献变量之一；降水和植被变量在所选时空混合效应模型中并不显著。模型生成的登革热概率图显示了风险的地理分布，与海拔梯度明显吻合。本文的结果为未来的登革热研究提供了最大的益处。
