Unit 3 · Compare groups
Post hoc tests in SPSS: Tukey and Games-Howell
Choose and interpret post hoc tests after ANOVA in SPSS, including Tukey HSD and Games-Howell.
Quick answer: How do you run post hoc tests in SPSS?
- In the One-Way ANOVA dialog, click Post Hoc….
- Tick Tukey if variances are equal, or Games-Howell if Levene’s test is significant.
- Click Continue, then OK.
- In Multiple Comparisons, look for pairs with Sig. below .05.
Report it in APA style: Tukey tests showed that Seminar students scored higher than Lecture students, p = .002.
What are post hoc tests?
A significant ANOVA tells you that at least one group differs, but not which. Post hoc tests (“after this”) compare every pair of groups while controlling the overall risk of a false positive. They’re the standard follow-up to a significant one-way ANOVA with three or more groups.
When to use it
- After a significant ANOVA when you had no specific predictions about which groups would differ.
- Use Tukey HSD when variances are equal and group sizes similar.
- Use Games-Howell when Levene’s test is significant or group sizes are very unequal.
- Bonferroni is a simple, conservative alternative, useful when you only have a few comparisons.
- If you predicted specific comparisons in advance, planned contrasts are more powerful.
How it works
Every test at α = .05 has a 5% risk of a false positive (Type I error). With three groups there are three pairwise comparisons, so the chance of at least one false positive across the family rises to about 14%. Post hoc tests control this familywise error rate. Bonferroni divides α by the number of comparisons (.05 ÷ 3 = .017). Tukey uses the studentized range distribution to keep the familywise rate at 5% while losing less power. Games-Howell does the same without assuming equal variances.
Assumptions
- The same assumptions as the ANOVA itself.
- Tukey assumes equal variances; Games-Howell doesn’t.
Worked example
After F(2, 57) = 6.53, p = .003, Tukey tests showed Seminar students scored 11 points higher than Lecture students (p = .002). Online didn’t differ significantly from either.
Key terms
- Type I error
- Concluding there’s a difference when there isn’t (false positive).
- Familywise error rate
- The chance of at least one Type I error across a set of tests.
- Tukey HSD
- “Honestly significant difference” test for all pairwise comparisons.
- Planned contrast
- A specific comparison chosen before seeing the data.
Common mistakes
- Running post hoc tests after a non-significant ANOVA and cherry-picking a significant pair.
- Using LSD, which applies no correction at all.
- Using Tukey when Levene’s test is significant.
- Reporting p-values without saying which correction was used.
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.
* Tukey for equal variances, Games-Howell (GH) for unequal variances.
ONEWAY score BY method
/STATISTICS DESCRIPTIVES HOMOGENEITY
/POSTHOC=TUKEY GH ALPHA(0.05).Practice data
Download the file used in this lesson and run it yourself in SPSS: you should get the same results as the screens above.
- teaching.savExam scores by teaching method: Lecture, Seminar, Online (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.