---
title: "One-way analysis of covariance (ANCOVA)"
method_id: SM.DIF.PAR.COV.ANCOVA
family: Difference
version: 1.1.0
date_modified: 2026-06-18
canonical: https://clarus.ofrencber.com/library/ancova.md
source: Clarus method library
---

# One-way analysis of covariance (ANCOVA)

> A classical statistical method in the Clarus library (Difference family).
> Clarus selects, assumption-checks and runs this method on your data with a
> deterministic rule engine and real numerical libraries (scipy / statsmodels);
> it never generates numbers, citations or results.

**Also known as:** analysis of covariance; covariate-adjusted ANOVA; ANCOVA.

## Hypotheses

- **Null (H0):** mu_adj_{g1} = mu_adj_{g2} = ... = mu_adj_{gk} (population covariate-adjusted means are all equal)
- **Alternative (H1):** at least one covariate-adjusted group mean differs from the others

## When to use it

Use this method when your goal is to:
- compare group means adjusting for covariate
- compare 3plus independent groups controlling continuous confounder

## Data it expects

- **dependent:** 1 continuous
- **independent:** 1 nominal/2+ levels
- **covariate:** 1+ continuous
- **pairing:** independent

## Assumptions Clarus checks

- **normality** (gate: shapiro_wilk; severity: warning; on violation: warn_only)
- **variance homogeneity** (gate: levene; severity: warning; on violation: warn_only)
- **homogeneity of regression slopes** (gate: homogeneity_of_regression_slopes; severity: critical; on violation: reroute)
- **linearity** (severity: warning; on violation: warn_only)
- **independence** (severity: warning; on violation: warn_only)

## Effect size reported

- **partial eta squared** (Cohen (1988); Lakens (2013))

## Honest limitations

- ANCOVA is only valid when the covariate-outcome relationship has the same slope in every group; if slopes differ, the adjusted comparison is misleading.
- In non-randomized designs, adjusting for a covariate does not recover a causal effect and can even introduce bias if the covariate is itself affected by group membership.
- Report covariate-ADJUSTED (estimated marginal) means, not raw group means, when describing the group effect.
- A significant group effect is not causal unless groups were formed by random assignment.
- Large n can pair statistical significance with a trivial adjusted effect — read partial eta-squared.
- Results generalize only to the sampled population and to the observed range of the covariate.

## Primary sources

- Fisher, R. A. (1932). Statistical Methods for Research Workers (4th ed.). Oliver & Boyd. [origin of analysis of covariance].
- Huitema, B. E. (2011). The Analysis of Covariance and Alternatives (2nd ed.). Wiley. https://doi.org/10.1002/9781118067475
- Field, A. (2013). Discovering Statistics Using IBM SPSS Statistics (4th ed.), Ch.12. Sage.
- Tabachnick, B. G., & Fidell, L. S. (2019). Using Multivariate Statistics (7th ed.), Ch.8. Pearson.
- Lakens, D. (2013). Calculating and reporting effect sizes. Frontiers in Psychology, 4, 863. https://doi.org/10.3389/fpsyg.2013.00863

## How to cite this

To cite this Clarus method page (the page itself — for the method's own primary sources, see above):

**Plain text**

Clarus, "One-way analysis of covariance (ANCOVA)", version 1.1.0, Clarus method library, 2026. https://clarus.ofrencber.com/library/ancova.md

**APA**

Clarus. (2026). One-way analysis of covariance (ANCOVA) (Version 1.1.0) [Statistical method, Clarus method library]. Retrieved from https://clarus.ofrencber.com/library/ancova.md

**BibTeX**

```bibtex
@misc{clarus-ancova,
  author       = {Clarus},
  title        = {One-way analysis of covariance (ANCOVA)},
  howpublished = {Clarus method library},
  version      = {1.1.0},
  year         = {2026},
  url          = {https://clarus.ofrencber.com/library/ancova.md}
}
```

_No DOI is minted for Clarus method pages yet; this citation uses the canonical URL, version and date. Cite the primary sources above for the method's scientific provenance._

---

METHOD SM.DIF.PAR.COV.ANCOVA · VERSION 1.1.0 · UPDATED 2026-06-18 · SOURCE Clarus method library