The Hong Kong Polytechnic University · FACULTY OF STATISTICS

RS517 Chap.1 Research Questions, Models and Ethics

- one subject, every graph, every model, every mark
6 Chapters3-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 1 of 10 · RS517

Research Questions, Models and Ethics

Questions before methods

A research question must identify the population, exposure or intervention, comparator, outcome and time point closely enough that the future data can be imagined. Broad clinical interests become researchable only after the literature gap, variables and target relationship are stated.

Directional wording should reflect prior evidence rather than a wish for a convenient test.

Conceptual models as analysis contracts

A conceptual model separates exposures, outcomes, mediators, confounders and effect modifiers. Those roles determine measurement timing and adjustment.

Controlling a mediator changes the target from a total effect to a direct effect, while omitting a confounder can create or hide an association.

Ethics inside design

Autonomy, beneficence, non-maleficence and justice shape recruitment, consent, burden and access to benefit.

Scientific validity is also ethical: participants should not carry risk for a study whose question or measurement cannot produce useful knowledge.

Audit the question before choosing a method

A question becomes operational when it fixes the population, exposure or intervention, comparator where needed, primary outcome and time point.

Test the wording by asking what one row of the future dataset would contain and what contrast would answer the question. If several incompatible outcomes or populations still fit, narrow it.

Then distinguish the estimand from the procedure: the scientific target might be an average treatment effect, prevalence, association or account of experience, while several designs can estimate that target under different assumptions.

Give every measured variable a role

A conceptual model should explain why a variable is collected.

A confounder precedes exposure and opens an alternative route to the outcome; a mediator lies on the pathway through which exposure may work; an effect modifier identifies heterogeneity. Timing alone does not settle every role, so state the causal story and draw it. Adjusting for a mediator changes the target away from the total effect, while conditioning on a collider can create bias.

The diagram is therefore a decision record rather than decoration.

Connect ethics to scientific value

Consent, privacy and participant burden are inseparable from design quality. An extra measurement needs a worthwhile purpose, recruitment must avoid exploiting dependence, and data access should be limited to what the question requires.

A weak design can be unethical because participant effort cannot produce a credible answer. In a proposal, pair each burden with the information it supplies and each foreseeable risk with a practical safeguard.

State who may decline, how withdrawal operates and what happens to data already collected.

Revision drill

Take one clinical topic and write a descriptive question, a treatment-effect question and a qualitative experience question. For each, name the population, data unit, central comparison and strongest boundary. Add a mediator and confounder only where the causal story supports them.

This drill prevents the familiar error of carrying one preferred test into questions that require different evidence.

In this chapter

What this chapter covers

  • 01

    Clinical problems and answerable research questions

  • 02

    Population, intervention, comparator, outcome and time

  • 03

    Conceptual models and causal roles

  • 04

    Mediators, confounders and modifiers

  • 05

    Consent, burden and scientific validity

  • 06

    Question-to-analysis alignment

Worked example · free

Narrow a clinical observation into a study

Q [8 marks]. Practice note: the marks used here are not part of the university's published assessment scheme. A team sees faster improvement among patients receiving home-practice messages. Write an answerable question and identify a mediator and confounder.
The funnel narrows a clinical concern into named variables, a population and one answerable relationship before analysis begins.The funnel narrows a clinical concern into named variables, a population and one answerable relationship before analysis begins.Clinical problemVariables and populationAnswerable question
The funnel narrows a clinical concern into named variables, a population and one answerable relationship before analysis begins.
  • 4Specify population, intervention, comparator, outcome and time.
  • 4Separate mediator from confounder by timing and causal role.
Among adults beginning outpatient rehabilitation after knee surgery, does a daily structured practice message, compared with usual printed instruction, improve six-week function? Adherence is a plausible mediator because the message may change it. Baseline motivation is a confounder if it influences both message exposure and later function outside random allocation.
Sia tip — Draw arrows before choosing adjustment variables; timing usually exposes whether a variable is a mediator or confounder.
Glossary

Key terms

Research Question
A specific statement linking a defined population, variables and outcome relationship.
Conceptual Model
A diagram of expected causal and associative roles among study variables.
Mediator
A variable on the pathway through which an exposure affects an outcome.
Confounder
A prior cause or correlate of both exposure and outcome that can distort their association.
Scientific Validity
The design quality required for participant burden to have a credible knowledge return.
FAQ

Research Questions, Models and Ethics FAQ

When is a research question specific enough?

It is specific enough when a reader can identify the target population, variables, comparison, outcome and time, then imagine one row of the future dataset. If several incompatible designs or outcomes still fit the wording, the statement remains a topic rather than an answerable question.

Why draw a conceptual model before analysis?

The model states which variables are exposures, outcomes, mediators, confounders and modifiers. Those roles determine measurement timing and whether adjustment answers the total effect, a direct effect or a distorted question. A list of available variables cannot supply that causal logic after results appear.

How does ethics affect statistical planning?

Ethics affects whether the question is meaningful, burdens are proportionate, consent is genuine and recruitment is just. Statistical planning matters because an underpowered or badly measured study can expose participants without a realistic route to useful knowledge, making scientific weakness an ethical weakness too.

Study strategy

Exam move

Take one clinical interest and write three versions: a broad topic, an answerable question and a testable hypothesis. Draw a conceptual model for the final version, then explain aloud why each extra variable is a confounder, mediator or modifier. Finish by naming the ethical justification for every participant burden.

Working through Research Questions, Models and Ethics in RS517? Sia is AskSia’s AI Statistics tutor — ask any RS517 Research Questions, Models and Ethics question and get a clear, step-by-step explanation grounded in how RS517 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 24 of your subjects at The Hong Kong Polytechnic University, plus 1,000+ more Bibles across the AskSia library.
Sia - your RS517 tutor, unlimited, worked the way the exam marks it
The full 3-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works