Repeated-measures and Crossover Designs

Following the same subjects across conditions

Repeated-measures designs measure the same units under multiple conditions or over time. Crossover designs are a specialized form common in clinical trials: each participant receives successive treatments separated by washout periods. These approaches control between-subject variability and increase statistical power, but they require careful management of carryover, period, and sequence effects.

Core Concept and Definition

A repeated-measures design involves observing each participant under multiple conditions or at multiple time points. Each subject thereby serves as their own control, allowing individual differences to be separated from error variance. The crossover design extends this logic to a clinical context: participants are randomly assigned to different treatment sequences, and a washout period follows each treatment to allow any residual effect to dissipate. Data collected from the same individuals in this way form a far richer and more efficient source of information than independent-groups designs.

How It Works: Main Types and Steps

Repeated-measures designs take several forms. In a fully within-subjects arrangement every participant experiences all conditions; in mixed (split-plot) designs some factors are between-subjects and others within. The simplest crossover form is the two-period, two-treatment AB/BA scheme: half the participants receive treatment A then B, the other half receive B then A. The researcher first randomly assigns participants to balanced treatment sequences, then collects measurements in each period, specifies an adequate washout interval, and finally analyzes all periods together. The statistical model explicitly accounts for period, sequence, and carryover effects.

A Concrete Application Example

Suppose pharmaceutical researchers want to compare two analgesics. By choosing a crossover design instead of independent groups, each patient receives both drugs in different periods; differences in metabolism, age, or pain threshold between groups are removed from the equation. In educational research, repeated measures are equally common: student vocabulary scores might be measured across three weeks under three different teaching methods. In both examples, taking measurements from the same individuals prevents individual differences from masking the comparison and makes it easier to reach meaningful conclusions even with smaller samples.

Common Pitfalls and Good Practice Principles

Carryover is the most serious hazard: residual effects of the first treatment can contaminate measurement of the second. An insufficient washout period amplifies this risk, so washout duration must be justified using domain-specific criteria such as drug half-life or learning decay. Period effects (maturation or fatigue over time) and sequence effects (the advantage or disadvantage of receiving a treatment first) must also be addressed. The sphericity assumption is critical in repeated-measures ANOVA; it should be tested with Mauchly's test and, if violated, corrected with Greenhouse-Geisser or Huynh-Feldt adjustments. Participant dropout is especially problematic in multi-period designs, requiring careful missing-data management.

Key terms

Carryover Effect: The persistent residual effect of a treatment in one period on subsequent periods.
Washout Period: A rest interval between successive treatments to allow carryover effects to dissipate.
Period Effect: Systematic change over time unrelated to treatment, such as maturation or fatigue.
Sphericity Assumption: The requirement that variances of difference scores between all pairs of conditions are equal.
Sequence Effect: The systematic influence of the order in which treatments are administered on outcomes.