Interpretation of Statistical Significance Testing

Abstract

There is some controversy about the interpretation, or even the necessity, of tests of statistical significance. One source of confusion hindering the resolution of the controversy is the multiplicity of philosophies and paradigms behind various tests of statistical significance. This article discusses and compares three paradigms (Fisher, Neyman-Pearson, and Bayes) that have been the foundation for approaches to the analysis of statistical hypotheses. The basic assumptions, decision logic, and the nature of the statistical hypotheses tested are outlined for each paradigm, and comparisons among the three paradigms are made. Practical implications are discussed and recommendations are made to guide selection of a paradigm and associated analysis based on the particular research goals.