Compare the means of three or more independent groups. Enter raw scores or group summary statistics and get the full ANOVA source table, exact p value, eta squared, omega squared, Bonferroni post hoc comparisons, and an APA 7 formatted result.
Need Levene's test, normality checks, and Tukey HSD? ReliCheck Quanta runs one-way and factorial ANOVA natively on your Mac with full assumption checking, marginal means plots, and publication-ready APA tables. Free to try.
Get QuantaUse a one-way ANOVA when you measured one continuous outcome across three or more independent groups and want to know whether the group means differ. Typical examples: three teaching methods, four dosage levels, five school sites.
With exactly two groups, an ANOVA and an independent samples t test give the same answer, and F equals t squared. Use the t test, since it reports a direction and a confidence interval for the difference. If the same people were measured under every condition, you need a repeated measures ANOVA, which you can run in Quanta.
The source table splits the total variation in your outcome into two parts. Between groups is the variation explained by group membership. Within groups is the variation left over, the spread of scores inside each group. The F ratio divides the between-groups mean square by the within-groups mean square.
An F near 1 means group membership explains about as much as noise does. The further F rises above 1, the more the group means are spread apart relative to the scatter inside each group. Degrees of freedom are reported as a pair: k minus 1 for the numerator, N minus k for the denominator.
Both report the proportion of variance in the outcome associated with group membership. Eta squared is the simple ratio of between-groups sum of squares to total sum of squares, and it is biased upward, especially with small samples and many groups. Omega squared corrects for that bias and is the better estimate of the population effect.
Report whichever your field expects, and report it every time. A significant F with an eta squared of .02 is a real but tiny effect, and the p value alone will not tell your reader that. Conventional benchmarks put .01 at small, .06 at medium, and .14 at large.
A significant F says at least one group mean differs from at least one other. It does not say which. That is what post hoc tests are for, and running them is not optional if you want to say anything specific about your groups.
This calculator uses Bonferroni corrected pairwise t tests built on the pooled within-groups mean square, which is the most portable option and the easiest to defend to a reviewer. Bonferroni multiplies each raw p value by the number of comparisons, so it is conservative: with many groups it will miss real differences. Tukey HSD is more powerful for all pairwise comparisons and is available in Quanta.
The F test assumes independent observations, roughly normal residuals, and equal variances across groups. This page computes the test; it does not verify those conditions, and neither does the number it returns.
Equal variances matters most when group sizes are unbalanced. If your largest group variance is more than about four times the smallest and the group sizes differ, the p value here is not trustworthy, and you want Welch's ANOVA instead. Independence matters most when people are clustered inside classrooms, clinics, or teams, in which case a one-way ANOVA is the wrong model no matter what the p value says.
An APA 7 report gives the F statistic with both degrees of freedom, the exact p value with no leading zero, and an effect size. Group means and standard deviations belong in the text or a table. The calculator assembles the sentence for you; copy it and replace the group labels with your own.
ANOVA is easy to run and easy to apply to the wrong design. If your groups are nested, measured more than once, or unbalanced in ways you did not plan, a fixed-price method review will catch it before a reviewer does.
Quanta