An elliptic spatial scan statistic and its application to breast cancer mortality data in Northeastern United States
An elliptic spatial scan statistic and its application to breast cancer mortality data in Northeastern United States
复制标题
椭圆空间扫描统计及其在美国东北部乳腺癌死亡率数据中的应用
DOI：
--
复制
发表时间：
2003
期刊：
影响因子：
--
通讯作者：
L. Pickle
中科院分区：
文献类型：
--
作者：
M. Kulldorff;Lan Huang;L. Pickle
S i131椭圆。如果真正的簇形状是非圆形的，则椭圆形扫描窗口可以提供更高的功率，这通常是预期的情况。在这里，我们描述和说明使用椭圆空间扫描统计，并将其应用于乳腺癌死亡率数据在美国东北部。讨论了椭圆形状和角度的选择，并通过功率分析比较了圆形空间扫描统计量和椭圆扫描统计量。地理和网络监控的潜在爆发检测和优先级系统G。P. Patil，W. L.迈尔斯角Taillie和D. Wardrop Pennsylvania州立大学我们提出了一个空间扫描统计的版本，旨在解决基于圆的扫描的以下缺点：·圆只能捕获紧凑形状的簇。在许多应用中，簇可以具有非常不规则的形状。圆柱形区域可以在时空扫描中产生差的热点描绘。·基于圆的分区依赖于欧几里德距离，不适用于沿网络沿着定义的数据。·空间扫描统计产生热点的最大似然点估计，但不提供不确定性或可变性的评估。人们希望有替代合理的描绘热点表示为热点的信心集。我们的扫描统计版本采用了在镶嵌单元（或网络节点）上定义的响应率的上水平集（ULS）。ULS扫描统计的吸引人的特征包括·识别任意形状的聚类·候选热点的数据自适应分区·对网络上的数据的适用性·为热点提供点估计和置信集·使用热点成员资格评级来映射热点边界不确定性·计算效率·对离散和连续响应的适用性·识别任意形状的集群在时空域中，我们还提出了一个优先级的创新。它在于能够根据多个指标和利益攸关方标准，使用偏序集，对热点进行优先排序和排名，而不必将指标纳入指数。生物监视应用扫描统计与多个，不同的数据源霍华德S。Burkom和尤金埃尔伯特国家安全技术部，约翰霍普金斯应用物理实验室，沃尔特里德陆军研究所研究人员在国防部全球新兴感染系统（DoD-GEIS）试点系统，电子监视系统的早期通知，
S i131 ellipses. An elliptic scanning window could provide higher power if the true cluster shape is noncircular, which one would often expect to be the case. Here, we describe and illustrate the use of an elliptic spatial scan statistic and apply it to breast cancer mortality data in the northeastern United States. The selection of the elliptic shapes and angles are discussed and comparisons are made between circular spatial scan statistic and elliptic scan statistic by implementing power study. Geographic and Network Surveillance for Arbitrarily Shaped Hotspots—Next Generation of Potential Outbreak Detection and Prioritization System G. P. Patil, W. L. Myers, C. Taillie, and D. Wardrop Pennsylvania State University We present a version of the spatial scan statistic that is intended to address the following shortcomings of circle-based scans: • Circles are able to capture only compactly shaped clusters. In many applications, clusters can have a very irregular shape. Cylindrical zones can yield poor hot spot delineation in a space-time scan. • The circle-based zonation relies on Euclidean distance and is inappropriate for data defined along a network. • The spatial scan statistic yields a maximum likelihood point estimate for the hot spot, but provides no assessment of the uncertainty or variability. One would like to have alternative plausible delineations of the hot spot expressed as a hot spot confidence set. Our version of the scan statistic employs the upper level set (ULS) of the response rate defined over the cells of a tessellation (or over the nodes of a network). Attractive features of the ULS scan statistic include • Identification of arbitrarily shaped clusters • Data-adaptive zonation of candidate hot spots • Applicability to data on a network • Provision of both a point estimate and a confidence set for the hot spot • Use of hot spot membership rating to map hot spot boundary uncertainty • Computational efficiency • Applicability to both discrete and continuous responses • Identification of arbitrarily shaped clusters in the spatial-temporal domain We also present a prioritization innovation. It lies in the ability for prioritization and ranking of hot spots based on multiple indicator and stakeholder criteria, using partial order sets, without having to integrate indicators into an index. Biosurveillance Applying Scan Statistics With Multiple, Disparate Data Sources Howard S. Burkom and Eugene Elbert National Security Technology Department, Johns Hopkins Applied Physics Laboratory, Walter Reed Army Institute of Research Researchers working on the Department of Defense Global Emerging Infections System (DoD-GEIS) pilot system, the Electronic Surveillance System for the Early Notification of