Choose your reading route
eduKateSG · Why Mathematics? · Stepped-Wedge Cluster Trials: Sequences, Periods and Time Adjustment
Start with the question, audit the worked example, or jump to the assumptions and practice. Every chapter stays open for uninterrupted reading.
Understand the question
Define the quantity and follow the mechanism.Route 2 · Chapters 8–14
Reproduce the example
Audit the numbers one step at a time.Route 3 · Chapters 15–21
Check assumptions
See where design and calibration matter.Route 4 · Chapters 22–28
Explore applications
Connect the method to a concrete investigation.Route 5 · Chapters 29–42
Study and verify
Practise, inspect failure cases and read further.
Full chapter index · Practice and checking · Mathematics Learning Hub
Why is mathematics important when an intervention is rolled out to clusters in stages? A stepped-wedge cluster trial randomises schools, clinics or communities to different crossover sequences. Valid inference separates intervention change from calendar change while accounting for repeated measurements within clusters.
Did You Know? If every cluster starts in control and ends with the intervention, a simple before-and-after comparison cannot distinguish the programme from a secular trend. Period adjustment is central, not optional decoration.
Full chapter index
Chapters 1–7 · Understand the question
Chapters 8–14 · Reproduce the example
Chapters 15–21 · Check assumptions
Chapters 22–28 · Explore applications
Chapters 29–42 · Study and verify
- Final decision
- A spreadsheet workflow
- A coding workflow
- Applications and careers
- A student investigation
- Common misconceptions
- Connection to recent mathematics articles
- Guidance for students and parents
- Practice questions
- Frequently asked questions
- A reporting checklist
- The grid is the design
- A final wedge audit
- Useful next reading
Clusters cross from control to intervention at different scheduled times. By the end, every cluster may receive the intervention, but the order of rollout is randomised.
Operational capacity may allow only a few schools or clinics to start at once. A randomised rollout can be more credible than choosing early sites by convenience.
Suppose six clusters are assigned to three sequences across four measurement periods. Two clusters switch after period 1, two after period 2 and two after period 3.
Sequence 1 clusters receive intervention in periods 2–4, sequence 2 in periods 3–4, and sequence 3 in period 4 only.
Across 6 clusters and 4 periods there are 24 cluster-periods. The illustrative design has 12 control and 12 intervention cluster-periods.
| Sequence | Clusters | First intervention period |
|---|---|---|
| 1 | A, B | 2 |
| 2 | C, D | 3 |
| 3 | E, F | 4 |
Intervention observations occur later on average than control observations. If outcomes improve over time anyway, an unadjusted effect will be biased upward.
Policy changes, seasons, staff experience or technology can shift outcomes across every cluster. Include categorical or suitably modelled period effects in the primary analysis.
| Condition | Cluster-periods |
|---|---|
| Control | 12 |
| Intervention | 12 |
At intermediate periods, some clusters are exposed and others are not. Comparisons within the same calendar period help separate treatment from time.
Measurements within a school remain correlated across people and periods. Models may need cluster random effects, cluster-period effects or repeated-cohort correlations.
Different people may be measured in each period. Correlation arises through the shared cluster and possibly shared staff, not repeated individual outcomes.
The same participants may be followed repeatedly. This adds within-person correlation and risks ageing, recovery or testing effects.
Implementation may take time. A prespecified transition period can be excluded or modelled, but removing inconvenient early outcomes after seeing them is not valid.
Analyse according to the randomised schedule when that matches the estimand, while documenting delayed or incomplete implementation and running sensitivity analyses.
The CONSORT extension defines sequences, periods and crossover, and requires reporting of time adjustment, correlations, implementation and cluster flow. Read the PubMed record.
The guidance explains why an unadjusted stepped-wedge analysis resembles an uncontrolled before-and-after study and why secular trends should be shown. Read the BMJ paper.
Power depends on number of clusters, periods, people per cluster-period, ICCs, cluster autocorrelation, effect, missingness and sequence allocation.
With few distinct crossover times, treatment and time effects can be highly correlated. More observations cannot fully repair weak schedule variation.
Large schools can dominate person-weighted analyses. State whether the estimand targets the average individual or average cluster.
Clusters may change behaviour before their scheduled switch because training or publicity begins early. Measure and discuss pre-exposure contamination.
A programme may need weeks to work. Immediate step functions can misrepresent build-up; lag models require prespecification and adequate periods.
Once implemented, an intervention may be difficult to remove, making a stepped wedge attractive. That practical fact does not make the design automatically ethical or efficient.
Six schools adopt a new mathematics timetable in three waves. Exam familiarity and term progression affect all schools, so the effect must be compared within periods.
Record training date, actual launch, exposure strength and interruptions. The randomised rollout and the delivered intervention are related but different data.
Generate cluster effects, period trends, correlation, unequal sizes, delayed implementation and missing periods. Check bias, coverage, Type I error and power.
Plot each cluster over time, mark planned and actual crossover, show unexposed secular trends and compare adjusted with unadjusted estimates.
Reproduce 12 control and 12 intervention cluster-periods, add a rising time trend and demonstrate the bias from ignoring period.
A stepped wedge cannot identify treatment cleanly when timing is confounded, implementation is universal at once or only one sequence exists.
Publish the rollout diagram, sequence allocation, period definition, correlation structure, time-adjusted effect and implementation deviations.
Use this design when randomised staged rollout is feasible and scientifically justified. Protect enough clusters, periods and concurrent contrasts for credible inference.
Create a cluster-by-period grid with planned condition, actual exposure, participants, outcome summary, missingness and transition flags.
Simulate correlated cluster-period data. Unit-test crossover dates, period coding, lag rules, missing periods and the primary mixed model.
Implementation researchers, education planners, public-health teams and operations analysts use rollout mathematics to separate intervention from time.
Simulate six clusters with a steady secular improvement but no treatment effect. Compare naive before-after and period-adjusted estimates.
Everyone receiving the intervention does not prove it works, later outcomes are not automatically treatment outcomes, and a rollout order is not random unless actually randomised.
The cluster-trial guide explains ICC and design effects. The platform-trial guide explains concurrent calendar comparisons.
Students should colour a cluster-period grid and then add a separate row for calendar trend. Parents can ask whether a school rollout simply coincided with normal progress.
- Draw three sequences.
- Count cluster-periods.
- Add a secular trend.
- Delay one crossover.
- Compare planned and actual exposure.
### Does every cluster get the intervention? Often, but not always.
### Why adjust for time? Treatment exposure is tied to later periods.
### Is rollout order enough? It must be randomised and implemented credibly.
### Can effects lag? Yes; the model should address that.
Report rationale, clusters, sequences, periods, randomisation, planned and actual crossover, participant type, time effects, correlations, missingness, model and interval.
A clear sequence-by-period diagram reveals the evidence structure faster than a participant total.
Confirm the intervention effect is estimated from within-period randomised contrasts and not from time alone, with every implementation deviation recorded.
Continue through the Mathematics Learning Hub. Recalculate the central worked example, identify the assumptions and explain which conclusion the result can and cannot support.
Extended practice laboratory
Practice case 1: List clusters
Setup: Start from the article’s stepped-wedge design mechanism for case 1. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 2: List periods
Setup: Start from the article’s stepped-wedge design mechanism for case 2. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 3: Assign sequences
Setup: Start from the article’s stepped-wedge design mechanism for case 3. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 4: Mark crossover dates
Setup: Start from the article’s stepped-wedge design mechanism for case 4. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 5: Count control periods
Setup: Start from the article’s stepped-wedge design mechanism for case 5. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 6: Count intervention periods
Setup: Start from the article’s stepped-wedge design mechanism for case 6. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 7: Add transition period
Setup: Start from the article’s stepped-wedge design mechanism for case 7. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 8: Add secular trend
Setup: Start from the article’s stepped-wedge design mechanism for case 8. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 9: Adjust for period
Setup: Start from the article’s stepped-wedge design mechanism for case 9. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 10: Model clustering
Setup: Start from the article’s stepped-wedge design mechanism for case 10. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 11: Add cohort participants
Setup: Start from the article’s stepped-wedge design mechanism for case 11. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 12: Add repeated participants
Setup: Start from the article’s stepped-wedge design mechanism for case 12. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 13: Delay implementation
Setup: Start from the article’s stepped-wedge design mechanism for case 13. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 14: Analyse by schedule
Setup: Start from the article’s stepped-wedge design mechanism for case 14. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 15: Analyse actual exposure
Setup: Start from the article’s stepped-wedge design mechanism for case 15. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 16: Add contamination
Setup: Start from the article’s stepped-wedge design mechanism for case 16. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 17: Add seasonality
Setup: Start from the article’s stepped-wedge design mechanism for case 17. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 18: Add site drift
Setup: Start from the article’s stepped-wedge design mechanism for case 18. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 19: Estimate treatment effect
Setup: Start from the article’s stepped-wedge design mechanism for case 19. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 20: Estimate correlations
Setup: Start from the article’s stepped-wedge design mechanism for case 20. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 21: Simulate null
Setup: Start from the article’s stepped-wedge design mechanism for case 21. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 22: Simulate time trend
Setup: Start from the article’s stepped-wedge design mechanism for case 22. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 23: Audit rollout
Setup: Start from the article’s stepped-wedge design mechanism for case 23. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 24: Name limits
Setup: Start from the article’s stepped-wedge design mechanism for case 24. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 25: Write conclusion
Setup: Start from the article’s stepped-wedge design mechanism for case 25. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the estimand and design, verify the statistical identity, state one diagnostic, record units and compare one defensible alternative with clear reasoning. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.