Question

What is a effect size that measures the proportion of variance for a one-way between-subjects ANOVA? Which one overestimates the size of an effect in a population?

Answer 1

Before I describe η2 G in more detail, some general comments regarding limitations are in order. Not all investigations lend themselves to proportion of variance effect size measures or other standardized effect size measures. As Bond, Wiitala, and Richard argue, when measurement units (e.g., reaction time in seconds) are clearly understood, raw mean differences may be preferred. Moreover, η2 G assumes a traditional univariate ANOVA approach, as opposed to multivariate or multilevel approaches to designs that include repeated measures. Finally, η2 and similar measures are often omnibus tests, which some scholars regard as insufficiently focused to capture researchers’ real concerns. At first glance, the matter of effect sizes in an ANOVA context seems straightforward. ANOVAs partition a total sum of squares into portions associated with the various effects identified by the design. These portions can be arranged hierarchically, and the various effects can be tested sequentially.

ANSWERED

PLEASE RATE ME

POSITIVE

THANKS