The Ethics of Statistical Testing

Abstract

Signicance testing is the most widely used statistical tool for quantitative analysis in science and business. We want to explore in what sense signicance testing can help in making ethical decisions, and in what sense it may obstruct them. In the rst section of this paper, we analyze a simplied model of ethical decision-making, showing how consistency in the assignment of probabilities is a prerequisite for any consequentialist justication of our choices. In the second section we provide a short introduction to signicance testing and its two main interpretations. In the third section we point to inconsistencies in the actual practice of signicance testing. Finally, we discuss several proposals for a consistent use of statistical tests in practical decision-making.