Statistics Glossary → Sphericity assumption
Sphericity assumptionconcept
The assumption in repeated measures ANOVA that the variances of the difference scores between all pairs of levels of a within-subject factor are equal.
When the same people are measured under three or more conditions, the repeated measures F test assumes that the difference between every pair of conditions has the same variance. For example, the differences between times 1-2, 1-3 and 2-3 should have similar variances. If the factor has only two levels, there is just one difference, so the assumption holds automatically.
When sphericity is violated, the F test becomes too liberal and the Type I error rate exceeds α. The assumption is tested with Mauchly's test, but that test has low power in small samples and is oversensitive in large ones. So the Greenhouse-Geisser or Huynh-Feldt epsilon correction is often applied directly; these corrections shrink the degrees of freedom by multiplying them by ε. Options that do not require sphericity are the MANOVA approach (multivariate tests) and mixed models.
An everyday example
A drug's effect is measured in the same patients at months 1, 2 and 6. The change between months 1 and 2 hardly differs from patient to patient, while the change between months 1 and 6 is very large in some and close to zero in others. If the variances of the differences are this unequal, sphericity is violated and a correction is needed.
In MerQur
'Mauchly sphericity test' and 'Sphericity correction' (None, Greenhouse-Geisser, Huynh-Feldt) in Repeated Measures ANOVA; the 'Sphericity correction' option (Automatic, Greenhouse-Geisser, Huynh-Feldt, Lower bound, All) in Mixed-Design ANOVA (Split-Plot).
Analyses where this concept appears
Repeated Measures ANOVAMixed-Design ANOVA, Split-Plot
Related concepts
Mauchly's test of sphericityGreenhouse-Geisser and Huynh-Feldt correctionsWithin-subject (repeated-measures) factor