Statistical methods for effect size analysis

Abstract

Informing in the studies about the effect size of an intervention or the impact of the factor(s) about an outcome, allows better decision-making for the application of the results in clinical practice. This article presents different methods to analyze the effect size, which can be through direct or indirect statistical methods. Within the direct methods, there's the difference in means between groups and the difference of absolute or relative frequencies. Among the indirect methods, Cohen's "d" family (which are based on standard deviation values), the "r and R2" family, measures of association (e.g. OR, RR, HR), and impact measures (e.g. NNT) are shown. The decision to use any of these methods depends on the objectives of the study and the measuring scale that is used to assess the results, as well as the data distribution. In order to enhance the understanding of the methods described in this article, examples are included, and the need to include level of precision (e.g. confidence intervals) is highlighted, along with the clinical decision thresholds for a better interpretation.