METHOD GUIDE
One-Way MANOVA in SPSS
Choosing the Multivariate Test and the Follow-Up
SPSS prints four multivariate statistics for one hypothesis, and what follows a significant result matters as much as the choice among them. Settling the outcome set, the covariance check and the follow-up before the model runs keeps both decisions defensible.
IN SHORT
What the Multivariate Test Answers and What Follows It
A one-way MANOVA compares the group means of a set of related continuous outcomes, taken together, across the levels of one categorical factor, and it asks whether the groups differ on the combination before any single outcome is examined. To run a one-way MANOVA in SPSS, open Analyze, General Linear Model, Multivariate, enter the outcomes as Dependent Variables and the grouping variable as a Fixed Factor, request homogeneity tests and effect size estimates under Options, and fix the follow-up plan before reading the multivariate tests. OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research, carries out and reviews this analysis through an in-house PhD-level statistician.
ANALYSIS SUPPORT
Analysis Support for This Method
A multivariate test only shows that the groups differ somewhere, so the harder work is deciding what to do next. Where the outcomes can be examined one at a time, a single-outcome comparison is the simpler route.
MANOVA
A joint comparison of several outcomes carried out within your study design, with the covariance check, the choice of test statistic and the follow-up agreed for your data.
One-Way ANOVA
The comparison of one outcome across three or more groups, for when the outcomes are best treated as separate questions and no joint test is wanted.
BEFORE YOU RUN IT
Decisions the Design Must Settle First
The model is only as defensible as the choice of outcomes and the sample behind them, and four decisions are made before the dialog is opened.
Outcomes That Belong Together The outcomes should be conceptually related and moderately correlated. Near-duplicate measures add little, and unrelated ones are better served by separate analyses with their own adjustment.
Cases per Group Each group needs more cases than there are outcomes, and more is better. A small group with many outcomes makes the covariance matrix unstable and weakens every test that follows.
Multivariate Outliers A case can look ordinary on every outcome and still be extreme in combination. Mahalanobis distance flags such cases, and the decision to keep or exclude one is stated, not made silently.
A Follow-Up Plan Decide before looking whether univariate tests, planned contrasts or an examination of the discriminant functions will follow a significant result, because a plan made afterward is exploratory.
IN SPSS
How to Run One-Way MANOVA in SPSS
One-way MANOVA runs through the GLM Multivariate dialog of IBM SPSS, and only the settings that alter the multivariate output are listed. Labels can differ between versions.
- Arrange one row per case, with the outcomes in separate columns and the group in a coded column.
- Open Analyze, General Linear Model, Multivariate, and move the outcomes into Dependent Variables and the group into Fixed Factor(s).
- Under Options, request descriptive statistics, estimates of effect size and homogeneity tests.
- Under Post Hoc, if the factor has three or more levels, choose one method; SPSS runs it for every outcome, so read only the outcomes your plan names.
- Run the model and read Box’s M test in SPSS and then the multivariate tests, before any univariate table.
- Follow up the outcomes named in the plan, and keep the outlier check with the analysis record.
READING THE OUTPUT
Reading the Multivariate Tests Table and the Tables Around It
Four statistics appear for one hypothesis, and the row labeled with the factor, not the Intercept row, carries the group effect.
| Output element | What it tells you | What to check |
|---|---|---|
| Box’s Test of Equality of Covariance Matrices | Whether the covariance matrices of the outcomes are equal across groups | That it is read with the group sizes, since it reacts to non-normality and a stricter cutoff such as .001 is often used |
| Multivariate Tests, Intercept row | Whether the overall mean vector differs from zero | That it is ignored, because it answers a question nobody asked about the groups |
| Multivariate Tests, factor row | Pillai’s Trace, Wilks’ Lambda, Hotelling’s Trace and Roy’s Largest Root, each with F, degrees of freedom, Sig. and partial eta squared | That one statistic was named in advance; with two groups all four agree, and Roy’s Largest Root is the least robust |
| Levene’s Test of Equality of Error Variances | A variance test for each outcome across groups | That it is a screen only, with Games-Howell chosen for follow-up when spreads differ |
| Tests of Between-Subjects Effects | One univariate ANOVA per outcome, with Intercept and Corrected Model rows | That these are read as follow-up tests under a stated adjustment and not before the multivariate result |
| Multiple Comparisons | Post hoc group differences, computed separately for each outcome | That the correction is understood to apply within an outcome, not across all of them |
| Descriptive Statistics | Means, standard deviations and counts by group and outcome | That the direction of every difference is described from here, as the multivariate row gives none |
REPORTING
A Reporting Frame for the Multivariate Result and Its Follow-Up
The multivariate statistic comes first and the follow-up second, in the order the plan fixed before any table was opened.
Name the outcomes and the factor, then give the statistic chosen in advance with its degrees of freedom, for example Wilks’ Λ = ___, F(df1, df2) = ___, p = ___, partial η² = ___. The GLM output prints no interval for partial eta squared, so any interval comes from another program. Say how Box’s test was judged and how many cases each group held. Then report the follow-up as planned: the univariate tests at the adjusted significance level, means with standard deviations for each outcome by group, and, where post hoc comparisons are used, mean differences with adjusted intervals. State how many outcomes were followed up, and label any further comparison exploratory.
COMMON MISTAKES
Six Habits That Undermine a MANOVA Write-Up
A MANOVA write-up is usually queried for a choice made after the multivariate table was seen, or for a follow-up that ignores how many outcomes there are.
Smallest p Reported Picking whichever multivariate statistic gives the smallest p is an undeclared choice. One statistic is named in advance, and the others are mentioned only where they disagree with it.
Univariate Tests Unadjusted One ANOVA per outcome at .05 each, presented as a follow-up, multiplies the false-positive chance. The significance level is adjusted for the number of outcomes followed up, and it is named.
Box's Test Over-Read A significant Box's test matters more with unequal group sizes, and a non-significant one is no proof of equal matrices. The test is weighed together with the group sizes.
Outcomes Piled In Adding every available measure to gain power dilutes the model and uses up degrees of freedom. The outcomes are chosen from theory, and the choice is stated before the model is run.
Combination Read as Every Outcome A significant multivariate test shows that groups differ on the combination, not that each outcome differs. Calling every outcome significant from that row alone overstates the finding.
Intercept Row Quoted The Intercept row tests whether the overall means are zero, which is rarely a hypothesis. The factor row carries the group effect, and quoting the wrong row is an easy slip.
Model already run? Independent Statistical Review checks the choice of statistic, the follow-up and the wording of the conclusion before you submit.
FREQUENTLY ASKED QUESTIONS
Questions About One-Way MANOVA Output and Follow-Up
Once the Multivariate Tests table is on screen, the next test, the row to read and the scaling of the outcomes need answers before anything is written.
What are the options for MANOVA follow-up tests in SPSS?
The usual follow-ups are univariate ANOVAs at an adjusted significance level, or an examination of the discriminant functions, which describes the combination of outcomes that separates the groups. The univariate results sit in the Tests of Between-Subjects Effects table, and with three or more groups post hoc comparisons or planned contrasts then locate the differences. Univariate tests ignore the correlation among outcomes, so the choice depends on whether the question is about each outcome or about their combination.
Why is the Intercept row the first thing in my Multivariate Tests table?
SPSS tests every term in the model, and the Intercept row asks whether the overall mean vector is zero, which is not a group question. Skip it and read the row carrying the name of your factor. Quoting the Intercept row in a write-up is a common slip, and it can look like a strong result when it is only a test of the constant.
Do I need to standardize outcomes measured on different scales?
No. The multivariate test statistics are unchanged when an outcome is rescaled, so standardizing is not needed for the test. What shapes the result is the correlation among the outcomes and the size of the group differences relative to the spread. Standardized values can still be useful when the means are presented side by side.
Can OnlineSPSS plan the follow-up tests for my MANOVA?
Yes. An in-house PhD-level statistician can check that the outcomes and the sample suit a joint test, set the follow-up plan and its adjustment before results are read, and carry out the analysis or review one you have run. If only the tables and the wording are outstanding, Results Reporting & Interpretation covers that stage. The research question and the conclusions remain with you.
How do you protect my research data and confidentiality?
Your data and project details are handled confidentially, and project information is not published. Researcher names and institutions are not used publicly without permission. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as the dataset or the output you have, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Outcomes Chosen but
the Follow-Up Still Open?
Describe the outcomes, the groups and their sizes, and attach any tables you have run. The assessment covers whether a joint test fits and how the follow-up will be planned and reported.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014