A visitor's guide to effect sizes - Statistical significance versus practical (clinical) importance of research findings

Effect Sizes (ES) are an increasingly important index used to quantify the degree of practical significance of study results. This paper gives an introduction to the computation and interpretation of effect sizes from the perspective of the consumer of the research literature. The key points made are: 1. ES is a useful indicator of the practical (clinical) importance of research results that can be operationally defined from being "negligible" to "moderate", to "important". 2. The ES has two advantages over statistical significance testing: (a) it is independent of the size of the sample; (b) it is a scale-free index. Therefore, ES can be uniformly interpreted in different studies regardless of the sample size and the original scales of the variables. 3. Calculations of the ES are illustrated by using examples of comparisons between two means, correlation coefficients, chi-square tests and two proportions, along with appropriate formulas. 4. Operational definitions for the ESs are given, along with numerical examples for the purpose of illustration.