CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: How to detect zero inflation in biological dosimetry data: An exact test for the Poisson distribution Authors:  Amanda Fernandez-Fontelo - Universitat Autonoma de Barcelona (Spain) [presenting]
Pedro Puig - Universitat Autonoma de Barcelona (Spain)
Elizabeth Ainsbury - Public Health England (United Kingdom)
Manuel Higueras - Basque Center for Applied Mathematics (Spain)
Abstract: The goal of biological dosimetry is to estimate the absorbed ionising radiation dose by an overexposed individual using chromosomal damage. When radiation occurs, damage in DNA is randomly distributed between cells which may be unrepaired to form aberrations (dicentrics). The radiation dose received by an individual is estimated building dose-response-calibration curves through the number of scored dicentrics when human body cells are exposed to radiation. Being exposed to whole (WBI) or partial body irradiation (PBI) determines how to proceed to estimate the absorbed dose. The Manual of the IAEA proposes the u-test to decide whether the irradiation is WBI or PBI. However, this test can sometimes be inappropriate. We propose to use an exact goodness-of-fit test for the Poisson distribution based on the occupational problems. This test allows us to analyze the zero-inflation/deflation of the data, being able to detect PBI. It can be seen as a complement of the u-test, being useful when the dispersion is insignificant, but the number of zeros is anomalous. Data coming from accidents are analyzed. An R Shinny app is presented which performs this and other zero-inflation tests under the Poisson assumption.