PES6251 Chap.7 Randomised and Quasi-experimental Designs
Randomised and Quasi-experimental Designs
Defend the counterfactual
Randomised controlled trials use chance allocation to support baseline comparability, while quasi-experiments introduce an intervention without random assignment and defend the counterfactual through comparison groups, time patterns or modelling. Parallel trials separate people into conditions.
Crossover trials let participants receive multiple conditions but require reversible effects, comparable periods, controlled sequence and an adequate washout to prevent carryover.
Protect allocation at each stage
Sequence generation, allocation concealment and blinding solve different problems.
Chance creates the sequence, concealment stops foreknowledge from shaping enrolment, and blinding reduces expectation or measurement effects after assignment. Report each protection rather than using randomised as a blanket description. Preserve the comparison in analysis by retaining original groups and accounting for every participant.
Excluding non-adherent participants can select people by post-allocation motivation or tolerance and reintroduce confounding.
Make a quasi-experimental comparison credible
A before-after result can reflect maturation, regression to the mean, seasonal progression or another event occurring at the same time.
Existing comparison groups, several pre-intervention observations and phased implementation can strengthen the estimate. Inspect whether groups followed similar trends before introduction and record policy, staffing or competition changes.
The conclusion should explain why the comparison approximates what would have happened without the programme, then name the most serious reason that approximation could fail.
Use crossover only for reversible contrasts
Within-person comparison controls stable differences and can improve efficiency, but sequence and period effects become central.
Randomise or counterbalance order, justify the washout and standardise visits. Learning, adaptation or durable treatment effects make crossover unsafe because the second period cannot recover an untreated state. Analyse paired outcomes while checking whether response differs by period or sequence. The design choice follows intervention behaviour, not a preference for smaller samples.
What this chapter covers
- 01
Random selection versus assignment
- 02
Quasi-experimental counterfactuals
- 03
Parallel-group trials
- 04
Crossover sequence and washout
- 05
Allocation concealment and blinding
- 06
Intention-to-treat and cluster analysis
Worked example · free
Choose a crossover design carefully
- 3Explain why within-person comparison helps.
- 3Specify sequence and washout.
- 3Name conditions that would make crossover unsafe.
Key terms
- Randomisation
- Assignment by chance intended to distribute baseline characteristics across intervention conditions.
- Quasi-experiment
- An intervention study without random assignment that uses design or analysis to estimate a counterfactual.
- Crossover Trial
- A trial in which participants receive more than one condition across planned periods.
- Washout Period
- A separation intended to remove the earlier intervention's effect before a later crossover period.
- Intention-to-treat
- Analysis comparing participants according to original assignment regardless of adherence or crossover.
Randomised and Quasi-experimental Designs FAQ
When is a quasi-experiment useful?
It is useful when random allocation is impractical or unethical but an intervention and credible comparison can still be observed. Existing comparison groups, interrupted time series or phased implementation can strengthen inference. The study must explain what would have happened without the programme and which rival explanations remain.
Why does a crossover trial need a washout?
The first intervention may continue affecting the second period. A washout creates time for that effect to end, but duration must be justified from the intervention and outcome. If learning or adaptation is effectively permanent, no realistic washout solves carryover and a parallel design is safer.
What is the difference between allocation concealment and blinding?
Allocation concealment prevents recruiters or participants from knowing the upcoming assignment before enrolment, protecting who enters each group. Blinding operates after assignment and reduces expectation or measurement effects. A trial can succeed at one and fail at the other, so report both separately.
Assessment move
Create a decision table comparing parallel, crossover and quasi-experimental designs by allocation, carryover, sample efficiency and counterfactual strength. For each study, identify the allocation unit and verify that the analysis respects it. Finish every critique by explaining why the comparison estimates what would otherwise have happened.
Separate sequence generation, allocation concealment and blinding, then state the distinct bias each protection addresses. For crossover proposals, test reversibility, washout, period and sequence effects before accepting the efficiency advantage. For quasi-experiments, graph several pre-intervention observations, record concurrent events and inspect whether the comparison followed a similar earlier trend.
Account for every allocated participant before interpreting the contrast.
Working through Randomised and Quasi-experimental Designs in PES6251? Sia is AskSia’s AI Statistics tutor — ask any PES6251 Randomised and Quasi-experimental Designs question and get a clear, step-by-step explanation grounded in how PES6251 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
