Statistics Glossary → Omega squared (ω²)
Omega squared (ω²)concept
An ANOVA effect size that estimates the proportion of variance explained by an effect in the population, with less bias than eta squared.
Eta squared (η²) is the ratio of a factor's sum of squares to the total sum of squares and gives the variance explained in the sample. Because random differences in the sample also contribute to that sum of squares, η² systematically overstates the population value, and the bias is marked in small samples. Omega squared corrects for this contribution using the error mean square: ω² = (SS_A - df_A·MSE) / (SS_total + MSE).
ω² is always smaller than η², and the two converge as the sample grows. If the effect is very small, the formula can give a negative value, which is usually reported as 0. Epsilon squared (ε²) makes a similar correction and gives values very close to ω². Values of roughly 0.01, 0.06 and 0.14 are often called small, medium and large, but these thresholds should be read in the context of the field.
An everyday example
Among 30 students taught by three different methods, the method seems to explain 20% of the variance in exam scores (η² = 0.20). In such a small sample, however, chance alone produces some 'explanation'; computing ω² brings the figure down to about 0.14. This suggests the method probably has a more modest, but still large, effect in the population.
In MerQur
It is requested with the 'Effect sizes: ω², ε², Effect CIs' checkboxes in One-Way ANOVA and 'Effect sizes: ω², ε²' in ANCOVA; One-Way ANOVA gives confidence intervals from the noncentral F distribution. Two-Way ANOVA computes it with the 'Effect sizes: η²ₚ, η², ω²' checkboxes and the Effect Size Calculator with 'Analysis type: η²/ω² (ANOVA, 3+ groups)'.
Analyses where this concept appears
One-Way ANOVATwo-Way ANOVAANCOVAEffect Size Calculator
Related concepts
Eta squared (η²) and partial eta squared (η²ₚ)Epsilon squared (ε²)