One-Way ANOVA (Compare 3+ Group Means)
The member the t-test family was missing.
Example
You enter
- Groups (one group per line; values separated by spaces or commas) 88 90 92 85 91 79 82 80 78 84 93 95 91 90 94
You get
- F stat 32.236
- Effect size (eta squared) eta^2 = 0.843
Details, formula, and sources
Where a two-sample t-test compares two means, one-way ANOVA compares three or more at once. Paste each group on its own line (values separated by spaces or commas); it partitions the variation into between-groups SSB = sum n_i (mean_i - grand)^2 (df k-1) and within-groups SSW = sum (x - group mean)^2 (df N-k), then F = (SSB/(k-1))/(SSW/(N-k)) with the F-distribution p-value from the bundled special-function helper. Three classes scoring means of 89.2, 80.6, and 92.6 give F = 32.2 on 2 and 12 df, p = 1.5e-5 - the classes differ. Also reports eta^2 = SSB/SST (variance explained: 0.01 small, 0.06 medium, 0.14 large). Verified against scipy.stats.f_oneway. ANOVA flags THAT a difference exists, not which pair - a Tukey HSD post-hoc test locates it; assumes roughly normal groups with similar variances. A statistics aid; the study design governs.
SSB = sum n_i (mean_i - grand)^2 (df = k-1); SSW = sum sum (x - mean_i)^2 (df = N-k); F = (SSB/(k-1)) / (SSW/(N-k)); upper-tail p = I(x; df_w/2, df_b/2) the regularized incomplete beta with x = df_w/(df_w + df_b F); eta^2 = SSB/(SSB+SSW).
Per OpenIntro Statistics Chapter 7 (comparing many means with ANOVA), by name; the F-distribution p-value reuses the bundled regularized-incomplete-beta helper. Verified against scipy.stats.f_oneway.
Free at openintro.org; the F-test and its partition of variance are public-domain textbook statistics.
Estimate. AHJ and licensed professional govern.
Field names used by the API: groups_text, f_stat, eta_squared
- Variance partition SST = SSB + SSW; F = MSB/MSW with df k-1 and N-kOpenIntro Ch. 7
- F-distribution p-value p = I(x; df_w/2, df_b/2) the regularized incomplete beta, x = df_w/(df_w+df_b F)F-distribution CDF identity
- Scope flags that a difference exists, not which pair (needs a Tukey HSD post-hoc); assumes normal groups, equal variancesscope of this tile