Confidence Interval Estimations for Standardized Linear Contrasts of Means: The One-Way Fixed-Effects Between-Subjects Univariate Case

• In Event: Alternative Methods for Data Analysis

Abstract

We examine the sensitivity of four confidence interval (CI) estimation methods for effect size (ES) to nonnormality and unequal variances. This study aims to (1) apply and extend the noncentral, Bonett’s, the percentile bootstrap, and the bias-corrected and accelerated bootstrap method to constructing the CI for a standardized linear contrast of means; and (2) assess the performance of these methods under conditions specified by different (i) numbers of independent levels, (ii) population distributions, (iii) population variances, (iv) population ESs, and (v) sample sizes. This study is limited to (1) one-way fixed-effects between-subjects univariate designs, and (2) a single linear contrast of means. However, findings from this study will have implications for CI estimates for ES in more complex designs.

Authors

• Li-Ting Chen, Indiana University

• Chao-Ying J. Peng, Indiana University