JOSS/paper.md

In BenjaK/scanstatistics: Space-Time Anomaly Detection using Scan Statistics

title: 'scanstatistics: space-time anomaly detection using scan statistics' authors: - affiliation: 1 name: Benjamin Allévius orcid: 0000-0002-0927-7183 date: "2 May 2018" bibliography: paper.bib tags: - scan statistic - cluster detection - anomaly detection - spatiotemporal affiliations: - index: 1 name: Department of Mathematics, Stockholm University

Summary

The R package scanstatistics enables the detection of anomalous space-time clusters using the scan statistics methodology. Scan statistics are commonly applied in disease surveillance, where they are used to detect disease outbreaks as they emerge locally. In this setting, cases of a given disease are recorded continuously across a country, and are then aggregated spatially to (say) district level, and temporally to (say) weekly counts. Scan statistics accomplish the detection task by searching the recent records of clusters of neighboring districts for patterns that seem anomalous given either past counts or the counts outside the cluster currently searched.

The scanstatistics package implements several scan statistics, making it a partially overlapping complement to existing scan statistic software such as SaTScan. For example, the conditional Poisson [@Kulldorff2001] and space-time permutation [@Kulldorff2005] scan statistics are available in both SaTScan and scanstatistics, while only the latter implements scan statistics for zero-inflated data [@Allevius2018], count data with overdispersion [@Tango2011], an unconditional (expectation-based) Poisson scan statistic [@Neill2005], and a Bayesian scan statistic [@Neill2006].

The R package scanstatistics is available on CRAN and its source code is available on GitHub.

References

BenjaK/scanstatistics documentation built on May 5, 2019, 2:41 p.m.