SPSS command: SPSS UNIANOVA
Factorial ANOVA and ANCOVA (GLM Univariate) in your browser
GLM Univariate fits one scale outcome against up to four fixed factors and five covariates, which covers two-way and three-way ANOVA, ANCOVA and unbalanced designs.
Free · No account · Files stay on your device · Updated 18 October 2026
What it tells you
Crosstabs runs the SPSS UNIANOVA procedure on your device and reports the Tests of Between-Subjects Effects with Type I–IV sums of squares and partial eta squared, plus estimated marginal means, Levene's test and post hoc comparisons.
- Whether each factor, each interaction and each covariate has an effect on the outcome after adjusting for the other terms in the model.
- Whether an interaction is present — for example, whether the effect of ad format depends on the market — which a series of one-way ANOVAs cannot show.
- How much of the variance each effect explains, as partial eta squared, and how much the whole model explains (R² and adjusted R²).
- Covariate-adjusted group means (estimated marginal means) and Bonferroni- or Šidák-adjusted pairwise comparisons between them.
When to use it
Use it when
- Two or more categorical factors and one scale outcome, such as recall score by ad format and market.
- You need to control for a continuous covariate (baseline score, age, prior spend) when comparing groups — ANCOVA.
- Cell sizes are unequal and you need a defined choice of sums of squares: Type III is the default, Type I gives sequential tests.
Use something else when
- Only one factor and no covariates: one-way ANOVA gives the same F plus Welch and Brown-Forsythe tests.
- Several correlated outcomes analysed together: use GLM Multivariate (MANOVA).
- The same respondents measured more than once: use repeated-measures ANOVA or a linear mixed model.
- A binary, count or ordinal outcome: use logistic regression, generalized linear models or ordinal regression.
How to run it in Crosstabs
- 01
Open your data
Open a .sav, CSV or Excel file in the workspace and switch to Statistics mode. The file is parsed in your browser.
- 02
Choose GLM Univariate
In the General linear model group, pick GLM Univariate (UNIANOVA).
- 03
Assign the variables
Tick one Dependent variable, up to four Fixed factors and, for ANCOVA, up to five Covariates.
- 04
Specify the model
Under Model choose Full factorial, Main effects only or Custom (tick the effects to include), and under Sum of squares keep Type III (default) or choose Type I, II or IV.
- 05
Pick the extras
Estimated marginal means and Homogeneity tests are on by default; add Parameter estimates, Observed power or Lack of fit. Set Compare estimated marginal means to LSD (none), Bonferroni or Šidák, and tick any Post hoc tests for factors (observed means), including Waller-Duncan with its Type I/Type II error ratio.
- 06
Run analysis
Click Run analysis. Cases missing any selected variable are excluded listwise; if a weight is active, declare it as a frequency or a sampling weight first.
Worked example
Computed by the Crosstabs engine from the synthetic data below when this page was built — not typed in.
Data
48 synthetic respondents in a 2 × 3 design: ad format (Static, Video) crossed with market (North, South, West), 8 per cell. The outcome is an aided-recall score drawn from normal distributions with a fixed seed, with a built-in Video advantage and a small extra lift for Video in the South.
| Source | Type III Sum of Squares | df | Mean Square | F | Sig. | Partial Eta Squared |
|---|---|---|---|---|---|---|
| Corrected Model | 835.89 | 5 | 167.18 | 4.834 | .001 | 0.365 |
| Intercept | 155132.28 | 1 | 155132.28 | 4486.077 | <.001 | 0.991 |
| format | 514.83 | 1 | 514.83 | 14.888 | <.001 | 0.262 |
| market | 244.21 | 2 | 122.10 | 3.531 | .038 | 0.144 |
| format * market | 76.85 | 2 | 38.42 | 1.111 | .339 | 0.050 |
| Error | 1452.39 | 42 | 34.58 | |||
| Total | 157420.56 | 48 | ||||
| Corrected Total | 2288.28 | 47 |
R Squared = 0.365 (Adjusted R Squared = 0.290)
| Item | Levene Statistic | df1 | df2 | Sig. |
|---|---|---|---|---|
| Based on Mean | 0.899 | 5 | 42 | .491 |
| Based on Median | 0.230 | 5 | 42 | .947 |
| (I) market | (J) market | Mean Difference (I-J) | Std. Error | Sig. | 95% CI |
|---|---|---|---|---|---|
| North | South | −2.78 | 2.079 | .565 | −7.97 to 2.40 |
| North | West | 2.74 | 2.079 | .582 | −2.44 to 7.93 |
| South | West | 5.53* | 2.079 | .033 | 0.34 to 10.71 |
Based on estimated marginal means; Bonferroni adjustment. * Significant at the .05 level.
How to read it
Format gives F(1, 42) = 14.89, p < .001, partial η² = 0.262; market gives F(2, 42) = 3.53, p = .038; the format × market interaction gives F(2, 42) = 1.11, p = .339. The largest effect is format (partial η² = 0.262). Levene's test based on the mean has p = .491, so equal error variances are a reasonable assumption. The model explains R² = 0.365 of the variance in recall.
Output tables explained
- Between-Subjects Factors
- Each factor's levels with value labels and the number of cases per level, so you can spot sparse or empty cells.
- Levene's Test of Equality of Error Variances
- Levene's statistic based on the mean and on the median across all factor cells. A small p value warns that the error variances differ; it is shown only for models without covariates.
- Tests of Between-Subjects Effects
- The main table: sum of squares of the chosen type, df, mean square, F, significance and partial eta squared for the corrected model, intercept, every factor, interaction and covariate, with R² and adjusted R² in the footnote.
- Parameter Estimates
- B, standard error, t, significance and 95% confidence interval for each dummy-coded parameter; the last level of each factor is the redundant reference set to zero.
- Estimated Marginal Means and Pairwise Comparisons
- Model-based means for each factor (and the first two factors crossed), evaluated at the covariate means, with standard errors, confidence intervals, adjusted pairwise differences and a Univariate Tests F.
- Multiple Comparisons
- Post hoc comparisons of observed level means: LSD, Bonferroni, Šidák, Scheffé, Tukey HSD, Hochberg's GT2, Gabriel, Games-Howell, Tamhane's T2 and Dunnett's T3. Skipped for factors with fewer than three levels or when the model has covariates.
- Homogeneous Subsets (Waller-Duncan)
- Observed level means grouped into subsets whose members differ by less than the Waller-Duncan critical difference, which uses the factor's main-effect F, the error mean square and the harmonic-mean group size.
Assumptions and notes
- Cases are independent: each respondent contributes one row and appears in one cell.
- Residuals are roughly normal with equal variance across cells; Levene's test checks the variance part.
- ANCOVA assumes the covariate's slope is the same in every group; test this with a Custom model that includes the factor × covariate term.
- Type IV equals Type III when every cell is filled; with empty cells the panel falls back to Type III and says so.
- A sampling weight shows weighted descriptive statistics only; the F tests assume independent, frequency-weighted cases.
How the results are validated
The UNIANOVA engine is tested against statsmodels 0.15 (anova_lm Type I–III, OLS parameter estimates, pairwise_tukeyhsd), SciPy 1.18 (Levene, noncentral F for observed power) and R 4.6 with car 3.1.5 (Type II and III with sum-to-zero contrasts) and emmeans 2.0.4. Sums of squares agree to a relative 1e-9, F to 1e-8 and p to 1e-7; marginal means and their standard errors to 1e-10; post hoc p values to 1e-8, or 2e-6 for the studentized-range tests. Frequency weights are tested to reproduce the case-replicated analysis. Validation evidence
Questions
Which sum of squares type should I use?
Keep Type III for unbalanced designs with interactions: each effect is tested as if entered last, which is what most reports expect. Type I is sequential and depends on the order of terms; Type II tests main effects without the interactions that contain them.
Why are there no post hoc tests when I add a covariate?
Post hoc tests compare observed means, which ignore the covariate. With covariates the panel skips them and you should read the Pairwise Comparisons of estimated marginal means, which are covariate-adjusted.
What is the difference between estimated marginal means and the descriptive means?
Descriptive means are raw cell averages. Estimated marginal means come from the fitted model, weight every cell equally and hold covariates at their means, so they are the fairer comparison when cells are unbalanced.
How do I interpret partial eta squared?
It is the share of variance an effect explains once the other effects are removed: SS(effect) / (SS(effect) + SS(error)). Because each effect has its own denominator, partial eta squared values can sum to more than 1.
What about the legacy ANOVA command?
Run it here: the legacy ANOVA command's experimental and hierarchical sums of squares correspond to Type III and Type I in this panel.
Is my data uploaded?
No. The workspace reads the file and fits the model in a Web Worker in your browser; the data is not sent to Crosstabs servers.
Related procedures
- One-Way ANOVAONEWAY
- GLM Multivariate (MANOVA)GLM
- Repeated Measures ANOVAGLM WSFACTOR
- Linear Mixed ModelsMIXED
- Linear RegressionREGRESSION
- Bayesian One-Way ANOVABAYES ANOVA
- Free SPSS alternative: full coverage list
Run GLM Univariate on your own data
Open a .sav, CSV or Excel file. Everything runs in your browser; the file never leaves your device.
SPSS is a trademark of IBM Corp. Crosstabs is not affiliated with or endorsed by IBM.