Coefficient omega and alpha are both measures of the composite reliability for a set of items. Unlike coefficient alpha, coefficient omega remains unbiased with congeneric items with uncorrelated errors. Despite this ability, coefficient omega is not as widely used and cited in the literature as coefficient alpha. Reasons for coefficient omega's underutilization include a limited knowledge of its statistical properties. However, consistent efforts to understand the statistical properties of coefficient omega can help improve its utilization in research efforts. Here, six approaches for estimating confidence intervals for coefficient omega with unidimensional congeneric items were evaluated through a Monte Carlo simulation. The evaluations were made through simulation conditions that mimic realistic conditions that investigators are likely to face in applied work, including items that are not normally distributed and small sample size(s). Overall, the normal theory bootstrap confidence interval had the best performance across all simulation conditions that included sample sizes less than 100. However, most methods had sound coverage with sample sizes of 100 or more.
Keywords: binary; bootstrap; coefficient omega; composite reliability; confidence interval; dichotomous; interval estimate; nonnormality; ordinal; reliability.