SIGNIFICANCE TESTS AND EFFECT MAGNITUDE: REFLECTIONS AND PROPOSALS

Abstract

Several proposals that enable to complement the information offered in statistical hypothesis testing are described. Using these proposals reduce the most hard critic that significance tests have suffered: Significance tests do not offer information about the magnitude of the relationship among the involved variables. The proposals that are discussed in this paper are: Confidence intervals, effect size, binomial effect size display, counternull value and common language effect size indicator.