SO5005 Chap.1 Research Process and Study Designs
Research Process and Study Designs
Start with the claim
Research design links an answerable question to sampling, measurement, comparison and inference. Descriptive studies estimate what is present, observational analytical studies examine patterns without assigning exposure, and experiments assign an intervention under controlled conditions.
Cross-sectional, cohort and case-control structures organise time differently, so each permits a different claim.
Protocols make choices reviewable
A protocol fixes eligibility, recruitment, outcomes, timing and analysis before results can influence the question. It also makes burdens, consent and safeguards visible for ethical review.
Strong proposals state the intended conclusion first and justify every later design decision against it.
What this chapter covers
- 01
Questions and claims
- 02
Descriptive and analytical designs
- 03
Cross-sectional and longitudinal time
- 04
Experimental allocation
- 05
Protocols and ethics
- 06
Limits on causal inference
Worked example · free
Separate change from causation
- 2Identify the before-and-after comparison.
- 2Name plausible time-varying alternatives.
- 2Write a bounded conclusion.
Key terms
- Protocol
- A prespecified plan for who enters a study, what is measured and how evidence will be analysed.
- Cohort Study
- An observational design that follows exposure status toward later outcomes.
- Case-control Study
- An observational design that begins with outcome status and compares earlier exposure.
Research Process and Study Designs FAQ
Can a cross-sectional study establish cause?
A cross-sectional study can describe prevalence and association at one time, but exposure and outcome are measured together. Temporal order is therefore uncertain, and common causes may explain the pattern. A causal conclusion needs stronger design evidence than a single snapshot provides.
Why write the protocol before collecting data?
Prespecification prevents the observed results from quietly changing eligibility, outcomes or analysis. It also exposes the recruitment burden, privacy plan and safety controls to review. Deviations can still occur, but they should be documented and explained rather than hidden.
What makes one design better than another?
Fitness for the research question matters more than complexity. The preferred design can reach the intended population, measure the relevant variables, support the required comparison and produce a claim proportionate to its limitations. Feasibility and ethical burden also constrain the choice.
Assessment move
Take a health-research headline and rewrite its claim at three levels: description, association and causation. For each version, sketch the least elaborate design that could support it, then list the main alternative explanation and one measurement that would expose that alternative. Build a design notebook with four columns headed question, comparison, timing and permitted conclusion.
Place a new scenario in one row each day and refuse to name the design until those columns are complete. Next, draw the participant path from eligibility through recruitment, measurement and analysis. At every transition, write who could disappear and whether their absence changes the claim. Practise distinguishing a snapshot from follow-up by moving the outcome measurement on a timeline.
For a cohort design, mark exposure before outcome; for a case-control design, begin at outcome status and trace earlier exposure; for an experiment, identify who controls allocation. Add one sentence explaining why a more elaborate design would or would not be worth its burden.
Once the structure is stable, write a protocol paragraph containing eligibility, recruitment route, variables, measurement schedule, analysis and safeguards in that exact logical order. Swap the order deliberately and notice where the argument becomes hard to follow. Finish each exercise with a red-team sentence that gives the strongest non-causal explanation for the observed pattern.
Then write the evidence that would weaken that explanation. Before coursework submission, compare the conclusion with the first question word by word: a descriptive question should not end in a causal verdict, and an observational comparison should not acquire experimental language during editing. This routine trains the core course habit of making design limits visible before results are known.
Add a weekly transfer exercise using the same question under a different design. Compare what changes in eligibility, timing, measurement and conclusion, then write why the revised design does or does not strengthen causal interpretation. Keep a small vocabulary list of design terms, but attach each term to a decision rather than a definition.
Before presenting, ask a classmate to identify the comparison and permitted claim from your method paragraph alone. If they cannot, revise the paragraph until the design logic is visible without the title. Finally, prepare a one-minute defence of feasibility covering access to participants, measurement burden, available time and ethical review.
This forces ambitious designs to meet practical conditions and helps distinguish a theoretically strong design from one that cannot be carried out well. Archive the final diagram beside the proposal and revisit it after results are written; any new arrow or deleted population should be disclosed as a deviation rather than silently absorbed into the story. Label the reason, date and expected effect of each change.
Recheck the permitted conclusion afterward. Keep the boundary explicit for readers and future reviewers who must reproduce the reasoning.
Working through Research Process and Study Designs in SO5005? Sia is AskSia’s AI Statistics tutor — ask any SO5005 Research Process and Study Designs question and get a clear, step-by-step explanation grounded in how SO5005 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
