Statistics Glossary → Within-subject (repeated-measures) factor
Within-subject (repeated-measures) factorconcept
A factor whose levels are all measured on the same subjects, such as measuring the same people at different times or under different conditions.
In a between-subjects factor each subject belongs to only one group (e.g. female or male). In a within-subject factor each subject is measured at every level of the factor (e.g. before treatment, at month 1 and at month 6). This design removes stable person-to-person differences from the comparison; because each person is compared with themselves, the error variance shrinks and higher power is achieved with fewer subjects.
Measurements taken from the same person are not independent, so repeated-measures ANOVA requires the sphericity assumption (equal variances of the difference scores). Mauchly's test checks it, and when it is violated the Greenhouse-Geisser or Huynh-Feldt corrections reduce the degrees of freedom. A within-subject factor with only two levels cannot violate sphericity. Order and carry-over effects (learning, fatigue) should be balanced in the design. A design combining within- and between-subjects factors is called a mixed (split-plot) design.
An everyday example
A gym wants to measure members' endurance after three different training programmes. Instead of giving each programme to different members, every member tries all three in turn and is tested after each one. 'Programme' is a within-subject factor: each person acts as their own control, so a naturally fit member's advantage does not distort the comparison.
In MerQur
In Repeated Measures ANOVA, several within-subject factors are defined with the 'Within-subject factors (WSFACTORS)' parameter; in Mixed-Design ANOVA, 'Additional within-subject factor' models them together with between-subjects factors.
Analyses where this concept appears
Repeated Measures ANOVAMixed-Design ANOVA, Split-Plot