Eta Squared Calculator

Eta Squared Calculator

Turn an F statistic and its degrees of freedom into eta squared, omega squared and Cohen’s f — or start from the sums of squares. The one-way ANOVA calculator reports these too; this page is for when you only have F.

The F value from your ANOVA output, e.g. 13.5102.
Eta squared is the share of the total variance your grouping explains. A large F with few degrees of freedom can still be a small effect — always report the effect size next to the p-value.
ETA SQUARED · F(2, 12) = 13.51
0.6925

A large effect: 69.25% of the total variance is explained by the grouping, with ω² = 0.6252 and Cohen’s f = 1.5006.

Partial eta squared
0.6925
Omega squared
0.6252
Cohen’s f
1.5006
p-value
0.0008
F statistic
13.5102
Total N
15

Calculation

    APA reporting line

    F(2, 12) = 13.51, p = .001, η² = .69

    Where the effect sits on the magnitude scale

    The purple bar is η² = 0.6925; the lighter bar is ω² = 0.6252. The marks are Cohen’s cut-offs: 0.01 small, 0.06 medium, 0.14 large.

    Cohen’s guide (1988): η² of 0.01 is small, 0.06 medium, 0.14 large. Here η² = 0.6925 is well past large — the grouping explains most of the variance.

    An eta squared calculator turns an F statistic and its degrees of freedom into eta squared — the share of the total variance your grouping explains — plus omega squared and Cohen’s f, with an APA reporting line. The one-way ANOVA calculator reports these too; this page is for when you only have the F from an SPSS or R output.

    How to use the eta squared calculator

    Use From F and df when you have an F statistic and its two degrees of freedom, for example from a published ANOVA table. Use From sums of squares when you have SSB and SSW instead.

    1. Choose a mode: From F and df, or From sums of squares.
    2. Enter the F statistic and df1 and df2 — or SSB, SSW, df1 and df2 in sums-of-squares mode.
    3. Click Calculate, or just type: the results update as you type.
    4. Read the headline eta squared, then check the tiles for partial eta squared, omega squared, Cohen’s f and the p-value. The calculation ledger shows every step, the chart places the effect on Cohen’s magnitude scale, and the APA reporting line is ready to paste into your results section.

    Try it now: with F = 13.5102, df1 = 2 and df2 = 12 the calculator gives η² = 0.6925, ω² = 0.6252, Cohen’s f = 1.5006 and p = 0.0008. Switch to sums of squares and enter SSB = 40 and SSW = 60 (with df1 = 2, df2 = 12) and it gives η² = 0.40, ω² = 0.2857 and f = 0.8165.

    What is eta squared?

    Eta squared (η²) is the proportion of the total variance in your outcome that is explained by the grouping. In a one-way ANOVA it is simply the between-groups sum of squares divided by the total sum of squares:

    η² = SSB / SST

    It runs from 0 (the groups explain nothing) to 1 (the groups explain everything). It is the ANOVA equivalent of R² in regression: a standardised effect size that does not depend on sample size, so it says how big the difference is, not just whether it is significant. A p-value of 0.0008 tells you the groups differ; η² = 0.6925 tells you the grouping accounts for 69.25% of the variance — a very large effect.

    The F alone does not tell you the size. A huge F with a huge sample can be a tiny effect, and a modest F with a small sample can be a large one. Always report the effect size next to the p-value — most journals now require it.

    The three numbers: eta squared, omega squared and Cohen’s f

    The calculator reports three versions of the same effect, because different fields prefer different ones:

    • Eta squared (η²): SSB / SST. The share of variance explained. Easy to read, but slightly biased upwards — it counts chance variation as if the grouping explained it.
    • Omega squared (ω²): (SSB − df1 × MSW) / (SST + MSW). The same idea with the chance-level part removed, so it is the less-biased estimate of the population effect. It is always a little smaller than η².
    • Cohen’s f: √(η² / (1 − η²)). The effect on a standard-deviation scale, like Cohen’s d for two groups. Use it when you need an effect size for a power analysis. For a two-group comparison, the equivalent standardised effect is Cohen’s d.

    From F and df alone: η² = (F × df1) / (F × df1 + df2), ω² = df1(F − 1) / (df1(F − 1) + N) where N = df1 + df2 + 1, and f = √(F × df1 / df2).

    How big is the effect? Cohen’s magnitude guide

    Cohen (1988) suggested these benchmarks. They are rules of thumb, not laws — what counts as large depends on the field.

    MagnitudeEta squaredCohen’s f
    Small0.010.10
    Medium0.060.25
    Large0.140.40

    An η² of 0.6925 is well past the large cut-off; an η² of 0.40 is large too, and 0.01 is barely worth reporting. For the wider picture — what effect sizes mean, and why a p-value alone is not enough — see what is effect size?.

    Eta squared vs partial eta squared

    In a one-way ANOVA, eta squared and partial eta squared are the same number: partial η² = SSB / (SSB + SSW), and since SST = SSB + SSW, the two fractions are identical. They only differ in a factorial ANOVA, where partial η² leaves out the sums of squares of the other factors, so it is larger than η². If you are reporting a one-way ANOVA, either label is fine — but say which one you used.

    The APA reporting line

    The calculator builds the reporting line for you, for example: F(2, 12) = 13.51, p = .001, η² = .69. The F is given to two decimals, the p-value to three (with the leading zero dropped, as APA style requires), and η² to two. The line does not include a confidence interval for η²: that needs the noncentral-F inversion, and the ANOVA page does not report one either — so this page does not invent one.

    Worked examples

    Example 1 — from the ANOVA page (F and df)

    The one-way ANOVA on three groups of five gave F = 13.5102 with df1 = 2 and df2 = 12 (n = 15). From the F alone: F × df1 = 13.5102 × 2 = 27.0204, so η² = 27.0204 / (27.0204 + 12) = 27.0204 / 39.0204 = 0.6925. For omega squared, df1(F − 1) = 2 × 12.5102 = 25.0204 and N = 15, so ω² = 25.0204 / (25.0204 + 15) = 25.0204 / 40.0204 = 0.6252. Cohen’s f = √(27.0204 / 12) = √2.2517 = 1.5006. The p-value is 0.0008, and the APA line is F(2, 12) = 13.51, p = .001, η² = .69. The grouping explains 69.25% of the variance — a large effect by any standard. These values match the ANOVA page exactly.

    Example 2 — from sums of squares

    Suppose SSB = 40 and SSW = 60, with df1 = 2 and df2 = 12. Then SST = 40 + 60 = 100, so η² = 40 / 100 = 0.40. MSW = 60 / 12 = 5, so ω² = (40 − 2 × 5) / (100 + 5) = 30 / 105 = 0.2857. Cohen’s f = √(0.40 / 0.60) = √0.6667 = 0.8165. The F statistic is (40 / 2) / 5 = 20 / 5 = 4.00, with p = 0.0467, so the APA line is F(2, 12) = 4.00, p = .047, η² = .40. Partial η² is 40 / (40 + 60) = 0.40 — identical, as it must be in a one-way ANOVA.

    Edge cases

    • F = 6.0, df1 = 2, df2 = 12 (the ANOVA vs t-test example): η² = (6 × 2) / (6 × 2 + 12) = 12 / 24 = 0.5 — half the variance explained.
    • F = 0: η² = 0 and ω² = 0. The groups explain nothing. The omega-squared formula would give a negative number when F is below 1; a negative ω² has no meaning, so the calculator reports 0.
    • A non-significant F with a non-trivial η²: with a small sample, a real effect can fail to reach significance. Report the effect size either way — the p-value says whether the evidence is strong, η² says how big the effect is.
    • SSW = 0: if there is no within-groups variation at all, the sums-of-squares mode cannot compute an F — enter an F directly instead.

    Frequently asked questions

    What is the difference between eta squared and partial eta squared?

    In a one-way ANOVA they are the same number: both divide SSB by the total (SSB + SSW). In a factorial ANOVA they differ: partial eta squared divides SSB by SSB + SSW only, leaving out the sums of squares of the other factors, so it is larger. Report which one you used.

    Why is omega squared smaller?

    Eta squared is biased upwards: it counts chance variation as if the grouping explained it. Omega squared subtracts the part of SSB that chance alone would produce (df1 × MSW), so it is always a little smaller and is the better estimate of the population effect.

    Source: the formulas on this page are the standard definitions of eta squared, omega squared and Cohen’s f (Cohen, 1988, Statistical Power Analysis for the Behavioral Sciences, 2nd edition). All worked examples on this page were recomputed with Python (scipy.stats). Last updated 19 November 2026.

    See all 91 free statistics tools on Empirical Rule Calculators.