Effect size is commonly used in many research areas in the past decade. Also, confidence intervals associated with effect size are encouraged to be reported. In the previous study, the performance of confidence interval with Cohen’s d was investigated focus on two groups. The major purpose of this study was to extend more than two groups design and evaluate the performances of confidence intervals with omega-squared by three methods (parametric, percentile bootstrap, and bootstrap bias-corrected and accelerated confidence interval [BCA]). The results show that there are two major causes for the variation in coverage rates, accuracy of effect size estimation and width of the intervals. But an unexpected result was present in this study, the BCA method for omega-squared was generally not the best performer for nonnormal data.
Questions & Comments
1. The scenario of this study was similar to previous study so that many descriptions were excluded. Thus, it is hard to read when there are limited descriptions, i.e., how to define or calculate the width of CI.