Selected Indicators of Effect Size in Psychological Research
Arkadiusz Prajzner
Abstract
Reflecting the changing statistical practice in psychological research, dominated by null hypothesis testing using a decision about the level of significance of the results, the recommendations are indicated for reporting effect sizes in papers. The study presents the concept of the effect size and indicates its place in data analysis regarding to outcome’s significance. The purpose of the work is to describe selected effect size indicators and to point the need of use and their proper presentation and interpretation in social sciences empirical work data analysis reports. Considering statistical analysis approach limits based on significance level only, the study presents the possibility of including in the data analysis an indicator of a more practical use which is the size of the effect. By using the most popular analysis methods, such as, Student t -test, univariate analyses of variance in between- and within-group schemes as well as Wilcoxon test, Mann-Whitney’s U , Kruskal-Wallis H , Friedman’s test and considering analysis for qualitative data, matched to research plans indicators of the effect size were presented. The paper presents the use, calculation and interpretation of the size effect such as: Cohen’s d , Hedges g , delta, Glass’s r g , matched pairs correlation r c , eta-square, omega-square and epsilon-square, Kendall’s W and fi , Cramer’s V as well as odds ratio and relative risk. The presentation of the effect size indicators was contrasted with the corresponding research plans and the type of data collected.