2011 Fiscal Year Final Research Report

Fast numerical computation for multivariate distributions with combinatorial structure and its application to spatial epidemiology

Research Project

Project/Area Number	21500288
Research Category	Grant-in-Aid for Scientific Research (C)
Allocation Type	Single-year Grants
Section	一般
Research Field	Statistical science
Research Institution	The Institute of Statistical Mathematics
Principal Investigator	KURIKI Satoshi 統計数理研究所, 数理・推論研究系, 教授 (90195545)
Co-Investigator(Renkei-kenkyūsha)	TAKAHASHI Kunihiko 国立保健医療科学院, 技術評価部, 研究員 (50323259)
Project Period (FY)	2009 – 2011
Keywords	空間疫学 / スキャン統計 / コーダルグラフ / 多重検定
Research Abstract	In spatial epidemiology, it is important to evaluate the multiplicity-adjusted p-value of scan statistics by taking their spatial correlations into account. However, it is difficult to conduct exact numerical calculation because high-dimensional integrations are required. On the other hand, Monte Carlo simulations cannot yield precise results. In our proposed method, we represent the spatial correlations of scan statistics as a graph, and by extracting its Markov structure, rewrite the high-dimensional integration as a successive numerical integration. Although, the proposed method reduces computational time of integration drastically, it is not powerful enough for the real datasets. More research is needed for practical use.