FAQs
It compares two scaled variances. In a two‑sample setting it tests whether population variances are equal. In one‑way ANOVA it tests whether group means differ by comparing between‑group to within‑group variability.
For a conventional right‑tailed test, place the larger sample variance in the numerator so the statistic tends to the upper tail. The reported p‑values are tail‑aware and valid regardless of ordering.
Two‑sided tests whether variances are unequal in either direction. Right‑tailed tests if variance₁ > variance₂; left‑tailed tests if variance₁ < variance₂. Your hypothesis should reflect the scientific question.
Exact p‑values are computed from the F distribution via the regularized incomplete beta function, using numerically stable continued fractions and a Lanczos log‑gamma implementation.
Yes. Add two or more groups and paste numbers separated by commas, spaces, or semicolons. The tool computes SS, MS, F, p, and effect sizes (η², ω²).
Independent observations, approximate normality within groups, and (for ANOVA) equal variances across groups. When these fail seriously, consider robust or permutation methods.
For a two‑sided level (1−α) interval, bounds are [ (s₁²/s₂²) / F1−α/2(df₁, df₂) , (s₁²/s₂²) / Fα/2(df₁, df₂) ]. These use upper and lower F quantiles with df₁ = n₁−1 and df₂ = n₂−1.