Unit 3 · Compare groups
Welch's ANOVA and Games-Howell in SPSS
What to do when Levene's test is significant: run Welch's ANOVA and Games-Howell post hoc tests in SPSS and report them in APA style.
Quick answer: How do you run Welch's ANOVA and Games-Howell in SPSS?
- Click Analyze › Compare Means › One-Way ANOVA.
- Under Options… tick Homogeneity of variance test and Welch.
- Under Post Hoc… tick Games-Howell, then click OK.
- If Levene’s test is significant, report the Robust Tests table.
Report it in APA style: Anxiety differed between groups, Welch’s F(2, 34.61) = 6.26, p = .005.
Unequal variances: Welch’s ANOVA
Standard ANOVA assumes every group has roughly the same spread. When they don’t, and especially when group sizes also differ, its p-values can be wrong. Welch’s ANOVA is a version that doesn’t assume equal variances, and Games-Howell is its matching post hoc test.
When to use it
- Levene’s test is significant (p < .05).
- Group sizes are unequal and spreads look different.
- Many statisticians recommend reporting Welch routinely, because it loses almost nothing when variances are equal.
How it works
Welch’s test weights each group by its sample size divided by its variance, so noisy groups count for less, and adjusts the error degrees of freedom downwards. That’s why df2 usually isn’t a whole number. Games-Howell compares pairs using each pair’s own variances and its own df.
Assumptions
- Scale outcome, roughly normal within each group.
- Independent groups.
- It does not assume equal variances.
Worked example
Levene’s test showed unequal variances, F(2, 57) = 4.15, p = .021, so Welch’s ANOVA was used: F(2, 34.61) = 6.26, p = .005. Games-Howell tests showed the control group was more anxious than the mindfulness group (p = .010).
Key terms
- Homogeneity of variance
- Equal variances across groups.
- Levene’s test
- Tests whether group variances differ.
- Welch’s F
- ANOVA statistic corrected for unequal variances.
Common mistakes
- Ignoring a significant Levene’s test.
- Pairing Welch with Tukey instead of Games-Howell.
- Rounding Welch’s df2 to a whole number in the write-up.
How to do it in SPSS
The SPSS syntax
Prefer typing to clicking? Open File › New › Syntax, paste this in, select it and choose Run › Selection. Every dialog's Paste button writes the same kind of code.
ONEWAY anxiety BY group
/STATISTICS DESCRIPTIVES HOMOGENEITY WELCH
/POSTHOC=GH ALPHA(0.05).Practice data
Download the file used in this lesson and run it yourself in SPSS. Your results will closely match the screens above, though a few numbers may differ slightly.
- welch.savAnxiety in control, mindfulness and exercise groups (N = 60)
Now practise it interactively →
Reading is only half of it. Tap through the real SPSS menus, read the output and pick the right write-up in 6 quick questions, with instant feedback on every answer.
Practice for this lesson is part of Pro, with every unit, unlimited hearts and saved progress.