Premium
Scan statistics on graphs
Author(s) -
Marchette David
Publication year - 2012
Publication title -
wiley interdisciplinary reviews: computational statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.693
H-Index - 38
eISSN - 1939-0068
pISSN - 1939-5108
DOI - 10.1002/wics.1217
Subject(s) - scan statistic , statistic , outlier , window (computing) , graph , computer science , statistics , probability and statistics , summary statistics , mathematics , algorithm , combinatorics , operating system
Scan statistics are used in spatial statistics and image analysis to detect regions of unusual or anomalous activity. A scan statistic is a maximum (or minimum) of a local statistic—one computed on a local region of the data. This is sometimes called ‘moving window analysis’; in the Engineering literature. The idea is to ‘slide’ a window around the image (or map or whatever spatial structure the data have), compute a statistic within each window, and look for outliers—anomalously high (or low) statistics. We discuss extending this idea to graphs, in which case the local region is defined in terms of the connectivity of the graph—the neighborhoods of vertices. WIREs Comput Stat 2012 doi: 10.1002/wics.1217 This article is categorized under: Data: Types and Structure > Graph and Network Data Data: Types and Structure > Social Networks