Analyses

Repeated-measures ANOVA

Compare three or more measurements from the same people.

Edit on GitHub

When to use it

Use repeated-measures ANOVA when the same people contribute every score and you have three or more conditions in separate columns. For example: does a mood rating change across week 1, week 2, and week 3?

Two columns only: the paired-samples t-test is the simpler test. A between-subjects factor as well as the repeated one is mixed ANOVA.

Assumptions

The differences between each pair of conditions should have similar spread. That is sphericity. Tensr runs Mauchly’s test. If Mauchly’s p is below .05, the report says sphericity may be violated and adds Greenhouse–Geisser and Huynh–Feldt corrections.

Tensr does not run Shapiro–Wilk on the differences. If the scores are ranks or badly skewed, use the Friedman test.

Partial η² is computed for the within-subjects effect. You do not choose it in the dialog.

Running it in Tensr

Analyze → General Linear Model → Repeated Measures ANOVA. In chat: “Repeated-measures ANOVA on week1, week2, and week3 by id.”
Subject column identifies the person. Measure columns are the repeated scores, at least two. Each person should have one row.
There is no post-hoc option and no sphericity switch. Both the uncorrected test and, when Mauchly is significant, the corrections are printed.

Options

Prop

Type

Reading the output

Thirty-six people across three weeks. Later weeks were built a little higher, with enough noise that the weeks still overlap.

SourceFdf (num)df (den)p-value
condition8.278270< .001
ConditionnMeanSD
week13619.9065.479
week23620.8894.685
week33623.9534.962

Repeated measures ANOVA on 3 conditions, 36 subjects. Repeated measures ANOVA on 3 conditions, 36 subjects. Metrics: Subjects = 36; Partial η² = 0.191.

Reporting (APA 7)

Repeated measures ANOVA on 3 conditions, 36 subjects. Report the estimate in the table. This procedure is not summarised by one p-value.

Coming from SPSS

Analyze → General Linear Model → Repeated Measures.

SPSS prints tests of within-subjects effects with sphericity assumed, Greenhouse–Geisser, Huynh–Feldt, and lower bound in one table, plus optional pairwise comparisons. Tensr splits the uncorrected test, Mauchly’s test, and the two corrections into three tables, and does not run pairwise tests.

The rank alternative is the Friedman test. Two measurements: paired-samples t-test. A grouping factor as well: mixed ANOVA.