Online ANOVA Calculator (One-way, Two-way, Repeated Measures)

When to use this test

• •You have a continuous outcome and one categorical factor with three or more levels (one-way ANOVA).
• •You have two categorical factors and want to test main effects plus their interaction (two-way ANOVA).
• •The same subjects are measured under multiple conditions or time points (repeated-measures ANOVA).
• •You need to know whether any group differs from the others before running follow-up comparisons.
• •You want a single omnibus test instead of running many pairwise t-tests and inflating the type I error rate.

How to use it

1. 1

   Upload or paste your data

   Load your CSV or XLSX into the analyzer, or paste rows into the grid. Use long format: one column for the outcome, one column per factor, one row per observation.

2. 2

   Pick the ANOVA variant

   In Hypothesis Tests, choose One-way, Two-way or Repeated-measures ANOVA. Each expects a slightly different column layout — the catalog tells you which roles to fill.

3. 3

   Assign variables

   Drag the outcome column into the Dependent slot and each categorical factor into its Factor slot. For repeated measures, also assign a Subject ID column so the software can match observations within participants.

4. 4

   Read the ANOVA table and post-hoc output

   The output shows sums of squares, degrees of freedom, mean squares, the F-statistic and a p-value for each factor (and interaction if applicable). When the omnibus test is significant, DeepStats automatically runs Tukey HSD or Bonferroni to tell you which specific groups differ.

5. 5

   Export the report

   Download the ANOVA table as CSV, export the diagnostic plots, or copy the interpretation into your write-up. All reported statistics follow APA formatting conventions.

Example with sample data

A small one-way design with three teaching methods and an exam score outcome. Paste directly into the analyzer to reproduce the numbers below.

Method,Score
Lecture,72
Lecture,75
Lecture,78
Lecture,74
Online,82
Online,85
Online,80
Online,86
Hybrid,91
Hybrid,94
Hybrid,89
Hybrid,92

The analyzer returns F(2, 9) ≈ 40.3, p < 0.001, eta-squared ≈ 0.90. That is a huge effect — the teaching method explains ~90% of the variance in scores. The Tukey HSD table then shows that every pair of methods differs significantly, with Hybrid scoring highest.

How to interpret the results

You will see: F-statistic, degrees of freedom (between, within), p-value, eta-squared and partial eta-squared, Tukey HSD or Bonferroni post-hoc comparisons.

F-statistic

F = MS_between / MS_within. Intuitively, it compares how much the group means differ to how much observations scatter within each group. Large F means the between-group differences dwarf the within-group noise.

Degrees of freedom

Reported as (df_between, df_within). df_between is the number of groups minus one; df_within is the total sample size minus the number of groups. These define the F-distribution used for the p-value.

p-value

The probability of seeing an F at least as large as yours if no group means actually differ. Below 0.05 is the usual threshold for rejecting the null. A significant omnibus F tells you something differs — post-hoc tests reveal exactly what.

Eta-squared (η²)

Proportion of total variance explained by the factor. Conventional benchmarks: 0.01 small, 0.06 medium, 0.14 large. In repeated-measures or multi-factor designs prefer partial eta-squared, which holds other factors constant.

Post-hoc comparisons

When the omnibus test is significant, DeepStats runs Tukey HSD by default (or Bonferroni if you prefer). The output shows every pairwise mean difference, adjusted p-value and confidence interval, so you can identify which groups differ without inflating the family-wise error rate.

Assumptions

• Normality of residuals.Within each group the outcome should be roughly normally distributed. ANOVA is fairly robust to departures, but for strongly skewed data consider the Kruskal-Wallis test (non-parametric one-way) or the Friedman test (non-parametric repeated measures).
• Homogeneity of variance.Groups should have similar variances. Levene's test is reported automatically. When it fails, switch to Welch's ANOVA, which does not assume equal variances.
• Independence.Observations must be independent of each other. If observations are nested (students in classes) or repeated (same subject over time), use a mixed model or repeated-measures ANOVA instead.
• Sphericity (repeated-measures only).Variances of the differences between all pairs of conditions should be equal. Mauchly's test is reported; when it fails DeepStats applies the Greenhouse-Geisser or Huynh-Feldt correction automatically.

Related calculators

Frequently asked questions