Skip to content

ANOVA Studio

Up to five groups. Drag a mean handle to move a whole group, or single points. F compares the spread of the means (SSB) with the spread inside groups (SSW). Below, pairwise differences with and without correction for multiple comparisons.

0255075100G1 · x̄ 42.3 · s 10.8G2 · x̄ 52.3 · s 7.9G3 · x̄ 62.7 · s 10.0
ANOVA Studio: G1 mean 42.3, SD 10.8; G2 mean 52.3, SD 7.9; G3 mean 62.7, SD 10.0. F = 7.84, p .004, η² = 0.466

ANOVA table

SSdfMSFp
Between1,460.92730.437.843.004
Within1,676.31893.13
Total3,137.120
Welch11.86.387.013

Welch (1951) F; df₂ = (k²−1)/(3Λ)

Pairwise comparisons (95 % CI of differences)

G1−G2-10.0 p .157G1−G3-20.4 p .003G2−G3-10.4 p .1360
Pairwise comparisons (95 % CI of differences): G1−G2 -10.0, p .157; G1−G3 -20.4, p .003; G2−G3 -10.4, p .136
Scores (editable)
#G1G2G3
1
2
3
4
5
6
7

One-way F, Welch F, Tukey HSD and Games–Howell from the engine (validated against SciPy/statsmodels; the studentised range is a port of R ptukey). Two-way SS are sequential, exact for balanced cells. The repeated-measures table and the Greenhouse–Geisser correction match base R (aov with Error strata, anova.mlm). Synthetic, seeded scores.

F = 7.84, p .004, η² = 0.466

Formula

F = MSB/MSW = [SSB/(k−1)] / [SSW/(N−k)]η² = SSB/SSTω² = (SSB − (k−1)MSW)/(SST + MSW)

The same F can hide very different pairwise pictures: one outlying group or a gradual trend. Correction widens every interval; that is the price of looking at all pairs.

Method: F, p: one-way ANOVA · Welch (1951) F; df₂ = (k²−1)/(3Λ) · Tukey HSD / Tukey–Kramer (studentised range) · Synthetic, seeded data. Illustrative, not an empirical finding.

Validation: F, Welch, Tukey, Games–Howell, two-way SS: SciPy/statsmodels; repeated measures and G–G: base R (tests/unit) · numerically validated