The misuse of statistics: concepts, tools, and a research agenda

Abstract

This paper presents concerns regarding misuse of statistics in scientific work, especially in biomedical research. The paper discusses what is meant by "misuse." It appears that misuse arises from various sources: degrees of competence in statistical theory and methods, honest error in the application of methods, egregious negligence, and deliberate deception (misconduct.) The incidence of error is partly due to a perceived need to meet artificial statistical criteria for acceptance of research reports for publication by journals. There has been no systematic research into the prevalence of misuse or its breakdown by type. Nonetheless, there are ways to encourage, or even to enforce, good statistical practice. These can be greatly supported by use of available statistical ethics documents. This article suggests lines of further research that could define the problem more explicitly and that might lead to additional corrective measures.