One-way ANOVA compares the means of three or more groups that differ in one factor, using \(F = MSB/MSW\); it assumes independence, normality and equal variances.
Hypotheses. \(H_0: \mu_1 = \mu_2 = \dots = \mu_k\); \(H_1\): at least one mean differs.
Procedure
- Compute group means \(\bar x_i\), grand mean \(\bar x\), and correction factor if using the shortcut method \(CF = (\sum x)^2/N\).
- Total: \(SST = \sum x^2 - CF\), df \(= N-1\).
- Between groups: \(SSB = \sum \dfrac{T_i^2}{n_i} - CF\) (\(T_i\) = group total), df \(= k-1\).
- Within groups: \(SSW = SST - SSB\), df \(= N-k\).
- \(MSB = SSB/(k-1)\), \(MSW = SSW/(N-k)\), \(F = MSB/MSW\).
- Reject \(H_0\) if \(F\) > \(F_\alpha(k-1, N-k)\) from the table.
Example of use: comparing average yield of wheat under three fertilisers, where fertiliser is the single factor.
Assumptions of one-way ANOVA
- Independence: observations within and between groups are independent (random sampling / random assignment).
- Normality: the response variable is approximately normally distributed in each group (check with Q–Q plot or Shapiro–Wilk test).
- Homogeneity of variance (homoscedasticity): all groups have equal population variances (check with Levene's or Bartlett's test).
- Measurement scale: the dependent variable is continuous (interval/ratio); the factor is categorical with two or more levels.
- Additivity: each observation = overall mean + group effect + random error, with errors independent and \(N(0, \sigma^2)\).
If normality or equal variances fail badly, use Welch's ANOVA or the non-parametric Kruskal–Wallis test.