Table of Contents
Eta squared (often symbolized as η²) is a critical effect size measure in statistics used primarily within the context of Analysis of Variance (ANOVA) models. This metric serves to quantify the strength of the relationship between the independent variable(s) (factors) and the dependent variable in an experimental or observational study. Unlike inferential statistics, which focus on whether an effect exists, Eta squared tells researchers how much of the total variance in the outcome variable is attributable to manipulation or differences in the predictor variables. Initially, it can be conceptualized simply as the ratio of the variance explained by a factor to the total variance observed in the study.
The definition often encountered in simpler models, such as those involving just two variables, states that Eta squared is calculated by squaring the measure of correlation between the two variables. However, its most robust and common application is in complex models involving multiple factors or levels. For instance, in a study examining how both gender and study habits influence test scores, Eta squared allows us to determine the proportion of the variation in test scores uniquely explained by gender, the proportion uniquely explained by study habits, and the proportion explained by their interaction. A higher resulting value represents a stronger, more substantive relationship between the tested variables.
Understanding this metric is vital because statistical significance alone (often determined by a p-value) only confirms the existence of an effect, not its practical importance or magnitude. Eta squared provides the necessary context, offering a standardized measure that is independent of sample size. When reporting research findings, particularly in fields relying heavily on experimental designs, including an effect size measure like Eta squared is considered best practice for ensuring transparency and enabling meta-analysis by future researchers.
Eta squared is a measure of effect size that is commonly used in ANOVA models.
The Role of ANOVA and Eta Squared
Analysis of Variance (ANOVA) is a powerful statistical test designed to compare means across two or more independent groups, dissecting the total observed variance in the dependent measure. Within the framework of ANOVA, Eta squared plays the crucial role of quantifying the practical importance of the effects uncovered by the analysis. It moves beyond the binary question of significance (is there an effect?) to address the quantitative question of magnitude (how large is the effect?). This allows researchers to gauge whether a statistically significant finding holds real-world relevance.
Specifically, Eta squared measures the proportion of the total variance in the dependent variable that is uniquely associated with a specific main effect or interaction effect identified within the ANOVA model. For example, if a study has two factors, Factor A and Factor B, Eta squared can be calculated separately for Factor A, Factor B, and the A x B interaction. The calculation essentially partitions the total variability into parts explained by the model components and residual unexplained variability.
The resulting value is inherently intuitive; if the Eta squared for Factor A is 0.30, it means that 30% of the total variability observed in the outcome measure can be directly attributed to the differences in the levels of Factor A. This makes it an essential tool for interpreting the complexity inherent in multifactorial experiments. Furthermore, it aids in comparing the relative influence of different variables within the same study, offering a clear hierarchy of importance among the predictors.
Deconstructing the Formula: Sum of Squares Explained
The calculation of Eta squared relies fundamentally on the concept of Sum of Squares (SS), which is the cornerstone of ANOVA. The Sum of Squares represents variability in the data; by partitioning the total variability into different sources, we can determine the contribution of each factor. The basic formula is a ratio, comparing the variability explained by the factor of interest to the total variability observed across all data points.
The required components for the calculation are derived directly from the standard ANOVA summary table output. We need two main values to proceed: the Sum of Squares for the specific effect being analyzed (SSeffect), and the total Sum of Squares (SStotal) for the entire model. These values quantify the deviation of scores from their respective means, thereby capturing the raw spread of the data.
The total Sum of Squares (SStotal) is the aggregate variability present in the dependent variable and is calculated as the sum of all individual sums of squares: the SS for all main effects, the SS for all interaction effects, and the SS for the residuals (error). SStotal represents the maximum possible variability that could potentially be explained by the model. SSeffect, conversely, isolates only the portion of that total variability that is directly attributable to the factor under examination.
Step-by-Step Guide: Calculating Eta Squared
The formula used to calculate Eta squared is remarkably straightforward, emphasizing the simplicity of expressing the effect size as a proportion of total variability.
The core formula utilized is:
Eta squared = SSeffect / SStotal
where the components are defined as follows:
- SSeffect: This is the Sum of Squares specifically associated with the main effect or interaction effect currently being assessed. It represents the variability explained by that specific factor.
- SStotal: This represents the total Sum of Squares in the comprehensive ANOVA model, encompassing the variability explained by all factors plus the residual error (SStotal = SSfactors + SSresiduals).
The resulting value for Eta squared will always fall within the range of 0 to 1. Values approaching 1 indicate that the factor being analyzed accounts for nearly all of the observed variability in the dependent measure, suggesting a very strong relationship. Conversely, values closer to 0 imply that the factor explains little to none of the total variance, meaning the relationship is weak or non-existent, irrespective of the statistical significance found by the F-test.
Interpreting the Results: Rules of Thumb for Effect Size
While the calculated Eta squared value provides the exact proportion of variance explained, researchers require standardized guidelines to translate this numerical output into qualitative meaning (e.g., small, medium, or large effect size). These interpretive rules, often cited from seminal works in statistics, provide context for comparing results across different studies and disciplines.
It is important to note that these guidelines are contextual and may vary slightly depending on the specific field of study (e.g., psychology versus engineering). However, the general rules of thumb established by Cohen (1988) are the most widely accepted standard for interpreting the magnitude of Eta squared values, allowing for clear and consistent reporting of findings. These benchmarks help ensure that reported effects are not just statistically present, but also practically meaningful.
The standard conventions used to interpret the magnitude of Eta squared are as follows:
- .01: Corresponds to a Small Effect Size, indicating that 1% of the total variance is explained by the factor.
- .06: Corresponds to a Medium Effect Size, indicating that 6% of the total variance is explained by the factor.
- .14 or higher: Corresponds to a Large Effect Size, indicating that 14% or more of the total variance is explained by the factor.
Practical Application: A Weight Loss Study Example
To solidify the understanding of Eta squared, let us consider a practical research scenario. Suppose a team of health researchers is investigating the influence of two key factors—exercise intensity and gender—on the dependent variable of weight loss over a defined period. The primary goal is not only to confirm if these factors have an impact but also to determine which factor contributes most significantly to the variability in weight loss outcomes.
The study is set up as a two-factor experimental design. Researchers recruit a balanced sample of 30 men and 30 women (60 participants total). Participants of each gender are then randomly assigned into three different exercise programs: no exercise (control), light exercise, or intense exercise. The experiment runs for one month, after which the total weight loss achieved by each participant is recorded. This setup results in a 2 (Gender) x 3 (Exercise Intensity) factorial ANOVA design.
By utilizing ANOVA, the researchers are able to test three null hypotheses simultaneously: that there is no main effect of gender, that there is no main effect of exercise intensity, and that there is no interaction effect between gender and exercise intensity. The resulting output summarizes the data’s variability, which is necessary for calculating the effect size for each tested factor. The data below shows the summarized results of the ANOVA analysis, using exercise and gender as factors and weight loss as the dependent variable.
Analyzing the ANOVA Output and Calculations
The output table below provides the essential components needed to calculate Eta squared, specifically the degrees of freedom (Df), the Sum of Squares (Sum Sq), and the statistical test results (F value and p-value).
Df Sum Sq Mean Sq F value p value gender 1 15.8 15.80 9.916 0.00263 exercise 2 505.6 252.78 158.610 < 2e-16 Residuals 56 89.2 1.59
Before calculating the individual Eta squared values, we must first determine the SStotal, which is the total Sum of Squares across the entire model. SStotal is found by summing the Sum of Squares for all effects (gender and exercise) and the Sum of Squares for the Residuals (error term).
Calculating SStotal: SStotal = SSgender + SSexercise + SSresiduals. Using the values from the ANOVA table: 15.8 + 505.6 + 89.2 = 610.6. With the SStotal established, we can now proceed to calculate the Eta squared for each factor using the ratio SSeffect / SStotal:
- Eta squared for gender: 15.8 / 610.6 = .02588
- Eta squared for exercise: 505.6 / 610.6 = .828
Why Eta Squared Matters: Beyond the P-Value
The calculated Eta squared values provide crucial interpretative insights that the p-value alone cannot offer. For the gender effect, the Eta squared is approximately 0.026. Comparing this to Cohen’s guidelines, this falls between a small (0.01) and medium (0.06) effect size. In contrast, the Eta squared for exercise is 0.828, representing an extremely large effect size, explaining over 82% of the total variability in weight loss.
These results highlight a critical difference in interpretation. The p-value for gender (0.00263) is statistically significant (p < 0.05), suggesting that gender does have a non-zero effect on weight loss. However, the associated effect size (.026) indicates that this significant effect is quite small in practical terms. Conversely, the p-value for exercise (< 2e-16) is highly significant, and its massive effect size (.828) confirms that exercise intensity is the overwhelmingly dominant predictor of weight loss in this study.
A statistical significance test, yielding a p-value, can only confirm whether a significant association exists between variables. It is solely the measure of effect size, such as Eta squared, that provides insight into the practical strength and magnitude of that association. Reporting both statistics is essential for a comprehensive and meaningful summary of research findings, preventing misinterpretation of statistically significant but practically minor effects.
Refinements: Introducing Partial Eta Squared
While Eta squared ( η²) is the most fundamental effect size measure in ANOVA, researchers often encounter a related, and sometimes preferred, metric: Partial Eta squared ( η²p). The distinction lies in how they define the denominator (SStotal). Standard Eta squared uses the total variability of the entire experiment (SStotal = SSeffects + SSerror).
Partial Eta squared, on the other hand, calculates the proportion of variance attributable to an effect after excluding the variance associated with other factors in the model (but including the error term). The formula for Partial Eta squared is SSeffect / (SSeffect + SSerror). This measure is particularly useful in complex factorial designs because it isolates the effect’s influence relative only to the remaining unexplained variability, allowing for comparison across studies with differing numbers of factors.
Although standard Eta squared is preferred when the goal is to describe the effect size relative to the entire population variance, Partial Eta squared often yields larger values, especially in models with many factors, and is thus frequently reported in academic literature. Researchers must be careful to specify which version is being used when communicating results to avoid confusion regarding the true magnitude of the effect.
Cite this article
stats writer (2025). What is Eta Squared?What is the definition of Eta Squared?What is an example of Eta Squared?. PSYCHOLOGICAL SCALES. Retrieved from https://scales.arabpsychology.com/stats/what-is-eta-squaredwhat-is-the-definition-of-eta-squaredwhat-is-an-example-of-eta-squared/
stats writer. "What is Eta Squared?What is the definition of Eta Squared?What is an example of Eta Squared?." PSYCHOLOGICAL SCALES, 15 Dec. 2025, https://scales.arabpsychology.com/stats/what-is-eta-squaredwhat-is-the-definition-of-eta-squaredwhat-is-an-example-of-eta-squared/.
stats writer. "What is Eta Squared?What is the definition of Eta Squared?What is an example of Eta Squared?." PSYCHOLOGICAL SCALES, 2025. https://scales.arabpsychology.com/stats/what-is-eta-squaredwhat-is-the-definition-of-eta-squaredwhat-is-an-example-of-eta-squared/.
stats writer (2025) 'What is Eta Squared?What is the definition of Eta Squared?What is an example of Eta Squared?', PSYCHOLOGICAL SCALES. Available at: https://scales.arabpsychology.com/stats/what-is-eta-squaredwhat-is-the-definition-of-eta-squaredwhat-is-an-example-of-eta-squared/.
[1] stats writer, "What is Eta Squared?What is the definition of Eta Squared?What is an example of Eta Squared?," PSYCHOLOGICAL SCALES, vol. X, no. Y, ص Z-Z, December, 2025.
stats writer. What is Eta Squared?What is the definition of Eta Squared?What is an example of Eta Squared?. PSYCHOLOGICAL SCALES. 2025;vol(issue):pages.