In recent years, there has been a shift from considering only results of statistical significance testing to the inclusion of measures of practical significance. In the comparison of treatment group means this has been the use of the standardized effect size, the effect size index, eta-squared, omega-squared, and the interclass correlation. Based on several Monte Carlo simulations using the null hypothesis distribution as the population, replications in each of 209 number of samples (k=2 to 12) and sample size configurations (n=5 to 500), it was clear that a large proportion of effect size measures exceed the commonly recognized standards by chance and that the chance values are predictable based on number of samples and sample sizes. This presentation includes mean effect sizes, variance of effect sizes, and proportions achieving a priori standards by chance. Equations are provided which may be used to predict mean effect sizes in given situations. Future research needs in this area are presented.
Learning Objectives: At the conclusion of this presentation, the participant should be able to: understand the basics of commonly used measures of effect size used in analysis of variance, including the standardized effect size, the effect size index (Cohen), eta-squared, omega-squared, and the interclass correlation; understand that these effect sizes vary systematically by chance depending on numbers of samples and sample sizes; and be able to apply provided equations to predict chance-determined values for these effect size measures so they can be used as standards for assessing observed effect sizes.
Keywords: Methodology, Biostatistics
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: None
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.