Home › Statistics Homework Help › SPSS Help › ANOVA in SPSS
Run it, read it, report it

How do I run ANOVA in SPSS, for one-way and factorial designs?

To run a one-way ANOVA in SPSS you open Analyze > Compare Means and Proportions > One-Way ANOVA, put your continuous outcome in Dependent List and your group in Factor, tick Descriptive and the homogeneity of variance test under Options, and pick Tukey under Post Hoc. For a factorial design you use Analyze > General Linear Model > Univariate instead. This page shows the exact clicks, which number in the output decides the result, and how to write it in APA.

✓Correct output, read the right row ✓Copy-paste APA reporting ✓Human statisticians, no AI

What this page covers

One-way ANOVAFactorial ANOVALevene's test Tukey post-hocWelch + Games-HowellBetween Groups F Partial eta squaredAPA reporting
Since 2014statistics specialists
Any testrun and interpreted
★★★★★rated by students
3 hrsfastest turnaround

ANOVA, the analysis of variance, tests whether the mean of a continuous outcome differs across the levels of one or more categorical factors. A one-way ANOVA answers a single question: do three or more unrelated groups differ on average? A factorial ANOVA adds a second factor and asks two more questions on top of that, one main effect per factor plus the interaction between them. This guide is part of our statistics homework help and sits under our wider SPSS homework help. It gives the exact menu path, points at the one value in each table that decides the result, and shows how to report it.

One version note that trips people up. In SPSS version 29 and later the old Compare Means menu is named Compare Means and Proportions, so One-Way ANOVA now lives under Analyze > Compare Means and Proportions > One-Way ANOVA. Factorial and repeated-measures designs stay under Analyze > General Linear Model in every recent version.

Which ANOVA do I need?

Match your design to the right procedure, then follow its steps below.

Your designYour dataProcedure
One factor, three or more unrelated groupsContinuous outcome, one grouping variableOne-way ANOVA
Two factors, unrelated groups, and their interactionContinuous outcome, two grouping variablesFactorial (two-way) ANOVA
One factor but group variances are unequalLevene's test is significantWelch's ANOVA with Games-Howell
The same people measured three or more timesRepeated measures, one groupRepeated-measures ANOVA

Not sure your data meets the requirements? Both guides below list the assumptions and the fallback to use when one fails.

One-way ANOVA with Tukey post-hoc in SPSS

Compares the means of one continuous outcome across three or more unrelated groups, for example an exam score across three revision methods.

Analyze > Compare Means and Proportions > One-Way ANOVA. Move the outcome into Dependent List and the group into Factor. Click Options and tick Descriptive and Homogeneity of variance test (this is Levene's test); while you are there, tick Welch so the resistant test is ready if you need it. Click Post Hoc and tick Tukey under Equal Variances Assumed and Games-Howell under Equal Variances Not Assumed. Continue, then OK.

Assumptions. A continuous outcome, three or more independent groups, no serious outliers, approximate normality within each group (Shapiro-Wilk in Analyze > Descriptive Statistics > Explore), and equal variances, which SPSS checks with Levene's test from the Options checkbox. If normality is badly broken, switch to the Kruskal-Wallis H test.

Step one, read Levene's test. The Test of Homogeneity of Variances table decides which post-hoc test is valid. Read the Based on Mean row. If its Sig. is above .05 the variances are equal, so the standard ANOVA and Tukey are fine. If its Sig. is .05 or below the variances are unequal, so you read the Welch row of the output and use Games-Howell instead of Tukey.

Test of Homogeneity of Variances (exam score)
Levene Statisticdf1df2Sig.
Based on Mean0.412227.666
Based on Median0.388227.682

Here Sig. = .666, above .05, so the variances are equal and Tukey is the correct post-hoc test.

Step two, read the ANOVA table. The ANOVA table holds the omnibus result. Read the Between Groups row for the F ratio, its two degrees of freedom (Between Groups df, then Within Groups df) and the Sig. A Sig. below .05 means at least one group mean differs from the others, though it does not yet say which.

ANOVA (exam score)
Sum of SquaresdfMean SquareFSig.
Between Groups85.5242.74.467.021
Within Groups258.3279.6
Total343.829

Step three, read the Tukey comparisons. Because the omnibus test is significant, the Multiple Comparisons table shows every pair of groups. An asterisk on the Mean Difference flags a pair that differs significantly, and the Sig. column gives the exact p-value for that pair. Each pair appears twice with the sign reversed, so read one direction and ignore the mirror.

Multiple Comparisons, Tukey HSD (dependent variable: exam score)
(I) Method(J) MethodMean Difference (I−J)Std. ErrorSig.
Method AMethod B−3.30*1.39.046
Method AMethod C−3.80*1.39.034
Method BMethod C−0.501.39.989

The asterisks show Method A differs from both Method B and Method C, while Method B and Method C do not differ from each other.

Report it (APA)A one-way ANOVA showed that exam scores differed significantly across the three revision methods, F(2, 27) = 4.467, p = .021, η² = .25. Tukey post-hoc comparisons showed that Method A scored significantly lower than Method B (p = .046) and Method C (p = .034), while Method B and Method C did not differ (p = .989).
Report it (APA), Levene's significantBecause Levene's test indicated unequal variances, a Welch's ANOVA was run. Exam scores differed significantly across the three methods, Welch's F(2, 17.6) = 5.12, p = .018. Games-Howell post-hoc comparisons showed Method A scored significantly lower than Method C (p = .022).

Common mistakes

  • Reporting only the omnibus F and stopping there, without the post-hoc comparisons that show which groups actually differ.
  • Using Tukey when Levene's test is significant, where Welch's ANOVA with the Games-Howell post-hoc is the correct choice.
  • Reading the Within Groups row instead of the Between Groups row for the F and Sig.
  • Running several separate t-tests across the groups instead of one ANOVA, which inflates the error rate.

Levene's failing, outliers, or unequal group sizes on your own dataset? Our statisticians will run the correct version and interpret it. Get a quote →

Factorial (two-way) ANOVA in SPSS

Tests two categorical factors at once, for example revision method and gender, and the interaction between them, on one continuous outcome.

Analyze > General Linear Model > Univariate. Move the continuous outcome into Dependent Variable and both grouping variables into Fixed Factor(s). Click Options (or Model in some versions) and tick Descriptive statistics, Estimates of effect size and Homogeneity tests. For pairwise follow-ups on a significant factor, use EM Means with a Bonferroni adjustment, or Post Hoc for a factor with three or more levels. Continue, then OK.

Assumptions. A continuous outcome, two independent categorical factors, no serious outliers, approximate normality of the residuals, and equal variances across all factor combinations, which Levene's test checks from the Options checkbox.

What to read. The Tests of Between-Subjects Effects table. Read the row for each factor and the interaction row, taking F, Sig. and Partial Eta Squared from each. Read the interaction first: a significant interaction means the effect of one factor depends on the level of the other, and the main effects are then interpreted with care. Ignore the Corrected Model, Intercept and Error rows for reporting.

Tests of Between-Subjects Effects (dependent variable: exam score)
SourceType III SSdfMean SquareFSig.Partial η²
Method78.4178.48.94.005.199
Gender9.819.81.12.297.030
Method * Gender45.7145.75.21.028.126
Error315.6368.77
Report it (APA)A two-way ANOVA found a significant main effect of revision method on exam score, F(1, 36) = 8.94, p = .005, partial η² = .20, no significant main effect of gender, F(1, 36) = 1.12, p = .297, partial η² = .03, and a significant method by gender interaction, F(1, 36) = 5.21, p = .028, partial η² = .13. Simple-effects tests showed the advantage of the method held for one gender only.

Common mistakes

  • Interpreting the main effects while ignoring a significant interaction, which changes what the main effects mean.
  • Reading the Corrected Model or Intercept row instead of the factor and interaction rows.
  • Leaving out the effect size, so the reader sees the F and p but not how large the effect is.

Partial eta squared: the effect size to report

A significant F tells you an effect exists; partial eta squared tells you how big it is.

What it is. Partial eta squared is the proportion of variance in the outcome explained by a factor, after the variance explained by the other factors is set aside. It runs from 0 to 1. The common benchmarks from Cohen are 0.01 for a small effect, 0.06 for a medium effect and 0.14 for a large effect.

Where to find it. In a factorial design, tick Estimates of effect size under Options in General Linear Model, and SPSS adds the Partial Eta Squared column to the Tests of Between-Subjects Effects table shown above. For a one-way design run through Compare Means and Proportions, the ANOVA table does not print it, so either read the ANOVA Effect Sizes table that SPSS version 27 and later adds below the ANOVA table, or run the same model through General Linear Model, Univariate to get partial eta squared directly. For a single-factor model, partial eta squared equals eta squared, which is the Between Groups sum of squares divided by the Total sum of squares.

When an assumption fails: the fallback

Most ANOVA marks are lost not on the software but on ignoring a broken assumption. Check the assumption, and when it fails, switch to the matched approach rather than reporting an invalid result.

AssumptionHow SPSS checks itIf it fails
Normality within groupsShapiro-Wilk and Q-Q plots in ExploreUse the Kruskal-Wallis H test
Equal variancesLevene's test (Homogeneity of variance)Read the Welch row and use Games-Howell
No serious outliersBoxplots in ExploreInvestigate, and report with and without the point
Independence of observationsStudy design, not a testUse a repeated-measures or mixed model

Reading ANOVA output: the mistakes that change the grade

Procedures on this page were checked against the IBM SPSS Statistics documentation, the UCLA statistical computing SPSS guides and the Kent State University SPSS tutorials.

Or hand the ANOVA to a statistician

When Levene's fails, the design is factorial, or the deadline is unforgiving, a specialist runs it in SPSS on your file and explains the output so you can defend it.

What one order includes: the right ANOVA chosen for your design, run in SPSS on your data, the assumption checks, the correct post-hoc test, the output tables, a plain-language interpretation, and the result written in your required reporting style. A syntax file is included so the analysis is reproducible.
1

Send your ANOVA task

Upload your data file, the assignment brief and your deadline, then get a free quote from support.

2

Pay securely

Approve the price and a statistician who works in SPSS every day starts right away.

3

Download the analysis

Get the output, interpretation and write-up in your account, with free revisions if you need them.

Vetted and safe

Real statisticians, and your data protected

Your work goes to a specialist who passed a subject-specific statistics test and a background check, and who works in SPSS every day. What we deliver is a model answer and study aid for reference, and many students use it to learn the method and check their own analysis.

Everything stays private. Your data and details are confidential, we never contact your school, payments are secure, and the work is original and plagiarism-checked.

Get ANOVA help you can trust →

Every order includes

Right ANOVA for the designAssumption checks Correct post-hoc testOutput and interpretation Syntax file includedFree revisions
★★★★★
What students say
★★★★★

"This company is wonderful. I was able to depend on them all semester for my Statistics assignments. Dependable with competitive prices. Would definitely use again!"

LMJ
Verified review
★★★★★

"Me and my friends were extremely pleased with the service, communication and efficiency that there experts have demonstrated. We truly appreciate services you provided."

Jacob C.
Verified review

ANOVA in SPSS FAQ

The F value is the ratio of the variance between the group means to the variance within the groups. A larger F means the group means are spread further apart than you would expect from chance alone. On its own it means little; you read it together with its two degrees of freedom and the Sig. column, and a Sig. below your alpha, usually 0.05, tells you at least one group mean differs.

When Levene's test is not significant and the group sizes are similar, Tukey's HSD is the standard choice for comparing every pair of means. When Levene's test is significant, the equal-variances assumption is broken, so you switch to Welch's ANOVA and use the Games-Howell post-hoc, which does not assume equal variances. Only run post-hoc tests when the overall ANOVA is significant.

Partial eta squared is an effect size that reports the proportion of variance in the outcome explained by a factor, after the variance explained by the other factors is removed. It runs from 0 to 1, with rough benchmarks of 0.01 small, 0.06 medium and 0.14 large. SPSS prints it in the General Linear Model output when you tick Estimates of effect size under Options.

A one-way ANOVA has one factor and tests whether the outcome differs across its groups. A factorial, or two-way, ANOVA has two or more factors and tests each main effect plus the interaction between them, which asks whether the effect of one factor depends on the level of the other. One-way ANOVA lives under Compare Means and Proportions; factorial ANOVA lives under General Linear Model, Univariate.

Use Welch's ANOVA when Levene's test of homogeneity of variances is significant, meaning the group variances are not equal. In the One-Way ANOVA dialog, click Options and tick Welch to get the resistant test, then pair it with the Games-Howell post-hoc instead of Tukey.

Yes. Send your data file and the assignment and a statistician runs the one-way or factorial ANOVA in SPSS, checks the assumptions, chooses the correct post-hoc test, reports the F, degrees of freedom, Sig. and effect size, and interprets it in full before your deadline, with free revisions.

Get your ANOVA done right

Hand it to a statistician and get accurate output and a clear interpretation before your deadline.

Get my free quote →

Related SPSS help

We use cookies to improve your experience. See our cookie policy.