CutL: an alternative to Kulldorff’s scan statistics for cluster detection with a specified cut-off level

  • Barbara Więckowska | basia@ump.edu.pl Department of Computer Science and Statistics, Karol Marcinkowski University of Medical Sciences, Poznan, Poland.
  • Justyna Marcinkowska Department of Computer Science and Statistics, Karol Marcinkowski University of Medical Sciences, Poznan, Poland.

Abstract

When searching for epidemiological clusters, an important tool can be to carry out one’s own research with the incidence rate from the literature as the reference level. Values exceeding this level may indicate the presence of a cluster in that location. This paper presents a method of searching for clusters that have significantly higher incidence rates than those specified by the investigator. The proposed method uses the classic binomial exact test for one proportion and an algorithm that joins areas with potential clusters while reducing the number of multiple comparisons needed. The sensitivity and specificity are preserved by this new method, while avoiding the Monte Carlo approach and still delivering results comparable to the commonly used Kulldorff’s scan statistics and other similar methods of localising clusters. A strong contributing factor afforded by the statistical software that makes this possible is that it allows analysis and presentation of the results cartographically.

Downloads

Download data is not yet available.

Author Biography

Barbara Więckowska, Department of Computer Science and Statistics, Karol Marcinkowski University of Medical Sciences, Poznan
 
Published
2017-11-06
Section
Original Articles
Keywords:
Space clusters, Cut-off level, High incidence rate, Spatial pattern detection, Binomial test for one proportion
Statistics
Abstract views: 1182

PDF: 170
HTML: 786
Share it

PlumX Metrics

PlumX Metrics provide insights into the ways people interact with individual pieces of research output (articles, conference proceedings, book chapters, and many more) in the online environment. Examples include, when research is mentioned in the news or is tweeted about. Collectively known as PlumX Metrics, these metrics are divided into five categories to help make sense of the huge amounts of data involved and to enable analysis by comparing like with like.

How to Cite
Więckowska, B., & Marcinkowska, J. (2017). CutL: an alternative to Kulldorff’s scan statistics for cluster detection with a specified cut-off level. Geospatial Health, 12(2). https://doi.org/10.4081/gh.2017.556