Evaluation and implementation of statistical methods to assess effects in count data

Abstract

As part of its statutory duties in the field of chemical safety, the German Environment Agency (UBA) is responsible for the environmental risk assessment of chemicals. This risk assessment is based on data from standardised ecotoxicological laboratory and field tests, which are generally described in Test Guidelines (TGs) published by the Organisation for Economic Co-operation and Development (OECD). The respective OECD TGs mention statistical methods for evaluating test results and are linked to the OECD Document No. 54 'Current Approaches in the Statistical Analysis of Ecotoxicity Data: A Guidance to Application' published in 2006. However, some of the statistical methods described in OECD No. 54 are outdated, and there is a lack of suitable methods for analysing non-normally distributed data, e.g. in aquatic mesocosms and field studies on soil organisms and arthropods.As part of an in-house research project (FKZ 3723 674010), UBA is currently revising OECD No. 54, which provides guidance on the statistical analysis of ecotoxicological data. During this process, shortcomings in the statistical analysis of ecotoxicological data were identified, particularly in higher-level tests (e.g. aquatic mesocosms and terrestrial field studies).The report submitted contains the results of the expert opinion conducted as part of the aforementioned in-house research project. Its main objective was to develop an R-package that improves the statistical analysis of Poisson-distributed data by employing alternative statistical methods, such as the Computational Approach Test based on the Closure Principle (CP-CAT) and other binomial test methods. The project was completed with the successful development of a publicly available R-package on the CRAN and GitHub platforms.This report details the work conducted to develop an R-package for Poisson-distributed data and also includes power analyses for several statistical methods.