A0857
Title: Permutation-based detection method for space-time interaction in infectious disease
Authors: Yuanyuan Yan - University of North Carolina at Chapel Hill (United States)
Feng-Chang Lin - University of North Carolina at Chapel Hill (United States) [presenting]
Pei-Sheng Lin - National Health Research Institutes (Taiwan)
Abstract: Space-time interaction is commonly defined as cases that are relatively near in space and close in time, a typical characteristic of infectious diseases. Hence, space-time interaction detection methods play a critical role in understanding the spread of contagious diseases by uncovering the dynamic process of spreading patterns. Hypothesis testing methods have been widely used to detect whether space and time interact in the spatiotemporal point process. The Knox test, a testing statistic based on the pairing method, is commonly used to assess the space-time interaction by counting pairs of events within defined spatial and temporal thresholds. Although several methods have been proposed to improve its testing characteristics, it remains unclear whether the testing procedure suits infectious diseases when the clustering is usually dense and imminent. The aim is to show that a conventional Knox test may not have adequate testing characteristics, such as inflated Type I errors and low statistical power. A permutation-based method is proposed that handles the dependence between case pairs in a fully nonparametric approach that maintains the dependence structure when deriving the distribution of the Knox statistics. The method is demonstrated through comprehensive simulation scenarios and applied to data on the spread of Dengue disease in Taiwan.
