Institute of Population Health Sciences, National Health Research Institutes, Miaoli, Taiwan.
Department of Mathematics, National Chung Cheng University, Minxiong, Taiwan.
Stat Med. 2022 Jan 15;41(1):146-162. doi: 10.1002/sim.9227. Epub 2021 Oct 20.
Identifying transmission of hot spots with temporal trends is important for reducing infectious disease propagation. Cluster analysis is a particularly useful tool to explore underlying stochastic processes between observations by grouping items into categories by their similarity. In a study of epidemic propagation, clustering geographic regions that have similar time series could help researchers track diffusion routes from a common source of an infectious disease. In this article, we propose a two-stage scan statistic to classify regions into various geographic clusters by their temporal heterogeneity. The proposed scan statistic is more flexible than traditional methods in that contiguous and nonproximate regions with similar temporal patterns can be identified simultaneously. A simulation study and data analysis for a dengue fever infection are also presented for illustration.
识别具有时间趋势的热点传播对于减少传染病的传播非常重要。聚类分析是一种特别有用的工具，通过将物品按照相似性分组到类别中，可以探索观察结果之间潜在的随机过程。在传染病传播的研究中，聚类具有相似时间序列的地理区域可以帮助研究人员追踪传染病从共同来源的扩散路径。在本文中，我们提出了一种两阶段扫描统计方法，通过时间异质性将区域划分为不同的地理聚类。与传统方法相比，所提出的扫描统计方法更加灵活，因为可以同时识别具有相似时间模式的连续和不相邻区域。还进行了模拟研究和登革热感染数据分析，以举例说明。
