MerQuR+Statistics Glossary

Statistics Glossary → Eta squared (η²) and partial eta squared (η²ₚ)

Eta squared (η²) and partial eta squared (η²ₚ)concept

Effect size measures in ANOVA that give the proportion of variance explained by an effect.

Eta squared is the effect's sum of squares divided by the total sum of squares: η² = SS_effect / SS_total. Partial eta squared removes the other effects from the denominator: η²ₚ = SS_effect / (SS_effect + SS_error). In a one-way ANOVA the two are identical; in multi-factor designs the partial value is larger, and partial values of different effects can sum to more than 1.

Both give the proportion of variance observed in the sample and overestimate the population value; in small samples ω² or ε² are less biased. Partial eta squared depends on the design: the same effect can come out differently when other factors are added, so generalized eta squared is recommended for comparing studies with different designs. By Cohen's rough benchmarks, about 0.01 is a small, 0.06 a medium and 0.14 a large effect.

An everyday example

The effect of three advertisements on purchase intention is measured. If 40 units of the total variability in intention scores (SS_total = 500) come from differences between the ads, η² = 40 / 500 = 0.08: about 8% of the score differences are explained by which ad was watched, and the rest by other person-to-person factors.

In MerQur

The 'Effect sizes: η²ₚ, η², ω²' checkboxes in Two-Way ANOVA; η² with its confidence interval (alongside ε² and ω²) in One-Way ANOVA and η²ₚ in ANCOVA; generalized η² in Repeated Measures and Mixed-Design ANOVA; partial η² for each multivariate test and each dependent variable in MANOVA; the rank-based η²_H in Kruskal-Wallis and η²_Q in Cochran's Q; and the η²/ω² (ANOVA) analysis type of the Effect Size Calculator.

Analyses where this concept appears

One-Way ANOVATwo-Way ANOVARepeated Measures ANOVAMixed-Design ANOVA, Split-PlotMANOVAANCOVAKruskal-WallisCochran's Q TestEffect Size Calculator

Related concepts

Omega squared (ω²)Generalized eta squared (η²G)Effect size