Effect size is assistance to a statistically significant testing, which provides information regarding the importance or the magnitude of influence of an intervention or a treatment on outcome variables. This study echoed the use of effect size and examined the 3 approaches (parametric, percentile bootstrap, and bootstrap bias-corrected and accelerated confidence interval, termed as BCA) to estimating confidence intervals (CIs) for omega-squared, which is often used in ANOVA models. The manipulated conditions included 4 population effect sizes, 8 different distributions of a dependent variable, 3 different numbers of groups, and 3 conditions of the number of independent variables and interactions between independent variables as well as group variance homogeneity. (The detailed manipulations were described in other papers.) The authors concluded that the nonnormal distribution of data degraded the coverage rate of parametric CI, and the parametric and percentile bootstrap yielded clearly higher coverage rates. On the other hand, the BCA generally produced narrower ranges of CIs compared with the percentile method, and it also provided better coverage than the parametric method did in several conditions. Thus, the authors suggested using the BCA approach to calculating CI of omega-squared in ANOVA models.
Here are some comments for the present study:
First, it seems that none of the 3 approaches performed equally well in coverage rates and CI width. The present study did not provide strong evidence to persuade readers to agree with the conclusion. It will be appreciated that if the authors described some details regarding the estimation of the 3 approaches and some possible reasons for the findings. Readers might be able to make a judgment about the use of the 3 CI calculation methods.
Second, it will be good to address the relations/connections between Cohen’s d and omega-squared and then maybe it will come out some possible explanations for the different performance of the 3 approaches in CI estimations of Cohen’s d and omega-squared.