How to report Welch's ANOVA and Games-Howell

Fewer worked examples exist for this than for standard ANOVA, so here is the full pattern.

The omnibus test

A Welch's one-way ANOVA was conducted because Levene's test indicated unequal variances, F(2, 24.13) = 8.47, p = .002, ω² = .28.

Why the degrees of freedom are decimal. Welch adjusts the denominator df to compensate for unequal variances. Reporting 24.13 rather than a whole number is correct. Do not round it to look tidier.

The pairwise comparisons

Games-Howell post-hoc comparisons indicated that group A scored significantly higher than group B (mean difference = 4.21, 95% CI [1.34, 7.08], p = .003), while the difference between A and C was not significant (mean difference = 1.02, 95% CI [-1.88, 3.92], p = .65).

Report mean differences and confidence intervals, not p-values alone. The interval tells your reader the size and precision of the effect.

Do not apply a further correction

Games-Howell already adjusts for multiple comparisons through the studentized range distribution. Adding Bonferroni costs power for no gain.

If the omnibus is significant but nothing pairwise is

Report the omnibus, state plainly that no pairwise comparison reached significance, and name the closest. See significant ANOVA but no significant post-hoc.

In Plotara

The Methods panel writes these sentences with your numbers in them, including the decimal degrees of freedom and the confidence intervals.