RS517 Research Methods & Data Analysis
RS517 Overview
- The Hong Kong Polytechnic University
- Semester 1, 2026/27
- Postgraduate
- Statistics and clinical research
What this course asks you to connect
RS517 treats research methods and data analysis as one chain of decisions. A specific question identifies the population, variables and time point. A conceptual model separates exposure, outcome, mediator and confounder. The design decides which alternative explanations can be excluded.
- Assessed by Four online quizzes worth 20%, a research-proposal presentation worth 40%, and a final written test worth 40%.
- Final format The published final written test mixes multiple-choice questions and short answers.
- Core method Match question, design, measurement and data structure before choosing a statistical procedure.
What you get in the full RS517 Exam Bible
- 10 chapters, 43 pages — every examinable topic
- 29 original practice questions, each with a model answer in the Oral presentation format
- 10 worked examples, solved the way RS517 marks
- 17 computed figures, redrawn from the course’s own numbers
- Exam-morning plan for the Oral presentation (40%) — what to revise last, where marks are lost
Free on this page: assessment table, chapter map, one worked example, key terms, FAQ. The rest is in the full exam bible.
How RS517 is assessed
| Component | Weight | Format |
|---|---|---|
| Online quizzes | 20% | Four quizzes; one submission for each |
| Oral presentation (research proposal) | 40% | Research proposal presentation |
| Final written test | 40% | Mix of MCQs and short answers |
The current course introduction publishes these three weighted components and they total 100%. The timetable places the PowerPoint deadline on 23 November 2026, presentation week from 30 November to 4 December, and the final written test on 14 December. Confirm live submission settings on Canvas.
Assessment structure
The published component weights total 100%.
What other students are asking about RS517
- Where can I find lecture notes or study guides for RS517 at PolyU?
- The RS517 study guide in AskSia Library covers Research Methods & Data Analysis in 10 chapters, with the assessment table, a chapter-by-chapter topic map, key terms and a worked example, free to read on this page.
- Is there a final exam in RS517?
- No final exam is listed in the offering this AskSia Library guide is built from; RS517 is assessed by Online quizzes 20%, Oral presentation (research proposal) 40%, Final written test 40%.
- What does RS517 cover on experimental designs and validity?
- In AskSia Library's RS517 guide, the Experimental Designs and Validity chapter explains random allocation, quasi-experiments, internal threats and transportability.
- What does RS517 cover on sampling and survey design?
- In AskSia Library's RS517 guide, the Sampling and Survey Design chapter explains target populations, frames, probability routes, non-response and item design.
- What does RS517 cover on descriptive statistics and missing data?
- In AskSia Library's RS517 guide, the Descriptive Statistics and Missing Data chapter explains distribution shape, centre, spread, outliers and missingness mechanisms.
Current RS517 dates
| Date | Item | Control |
|---|---|---|
| 23 November 2026 | Presentation PowerPoint deadline | Published in the current timetable and presentation guide. |
| 30 November to 4 December 2026 | Presentation week | Published in the current timetable and presentation guide. |
| 14 December 2026 | Final written test | Published in the current timetable. |
Dates are as published in the Semester 1, 2026/27 timetable and presentation guide. Confirm exact deadlines and submission settings in the live LMS.
What RS517 covers
Ten chapters connect a clinical research question to design, sampling, analysis, measurement and evidence synthesis.
Research Questions, Models and Ethics
answerable questions, conceptual models, causal roles and ethical alignmentQuestions 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.
- Clinical problems and answerable research questions
- Population, intervention, comparator, outcome and time
- Conceptual models and causal roles
- Mediators, confounders and modifiers
- Consent, burden and scientific validity
- Question-to-analysis alignment
Experimental Designs and Validity
random allocation, quasi-experiments, internal threats and transportabilityAssignment determines inference
True experiments randomly allocate interventions, quasi-experiments study intervention changes without random allocation, and observational studies measure exposures that were not assigned.
- True experimental and randomised designs
- Quasi-experimental comparisons
- Observational exposure designs
- Internal validity threats and controls
- Efficacy and effectiveness
- External validity and transportability
Sampling and Survey Design
target populations, frames, probability routes, non-response and item designThree populations
The target population is named in the question, the sampling frame is the operational list from which selection occurs, and the achieved sample supplies data.
- Target population, frame and achieved sample
- Simple and systematic random sampling
- Stratified and cluster sampling
- Convenience, quota and purposive routes
- Coverage and non-response error
- Survey item construction
Descriptive Statistics and Missing Data
distribution shape, centre, spread, outliers and missingness mechanismsDescribe before testing
The mean uses every value and is pulled by tails, while the median identifies the ordered middle and resists extremes.
- Variable scales and suitable summaries
- Mean, median and distribution shape
- Standard deviation and interquartile range
- Outliers and influence
- Missing-data patterns and mechanisms
- Complete-case, model-based and sensitivity approaches
Hypothesis Tests, Errors and Power
null models, confidence intervals, Type I and Type II errors, power and multiplicityA decision under a model
The taught sequence states the null, sets alpha, chooses and computes a statistic, obtains a p-value and compares it with alpha.
- Null and alternative hypotheses
- Alpha and the testing sequence
- Type I and Type II errors
- Power and sample information
- Confidence intervals and clinical thresholds
- Multiplicity and analysis families
Choosing Tests for Group Comparisons
outcome scale, independent and paired data, ANOVA and rank-based alternativesStructure before assumptions
Test selection begins with the question and outcome scale, then asks whether observations are independent or related and whether two or several groups are compared.
- Outcome scale and test families
- Independent and paired observations
- Independent and paired t tests
- One-way and repeated-measures ANOVA
- Rank-based comparison procedures
- Global tests and follow-up contrasts
Factorial and Multivariate Analysis
main effects, interactions, repeated measures, covariates and multiple outcomesInteractions before averages
A factorial model estimates main effects and interactions. A main effect averages across the levels of another factor, while an interaction shows that one contrast changes across those levels.
- Main effects and interactions
- Simple effects and profile plots
- Between-person and within-person factors
- Group-by-time interpretation
- ANCOVA and covariate roles
- MANOVA and MANCOVA
Correlation and Regression
linear association, conditional coefficients, prediction, residuals and diagnosticsAssociation and prediction
Pearson correlation summarises the strength and direction of a linear relation and is symmetric in its two variables.
- Linear association and scatterplots
- Pearson correlation and restricted range
- Regression equations and predictions
- Conditional coefficient interpretation
- Residuals, variance and influence
- Extrapolation and model validation
Reliability, Validity and Diagnostic Accuracy
measurement error, agreement, validity evidence, sensitivity, specificity and ROC reasoningConsistency is not correctness
Reliability concerns consistency across occasions, raters or forms. Validity concerns whether evidence supports the intended interpretation and use.
- Signal, random error and systematic bias
- Test-retest and rater reliability
- Correlation versus agreement
- Face, content, criterion and construct validity
- Sensitivity and specificity
- Predictive values and ROC thresholds
Evidence Synthesis and Research Inference
systematic review, meta-analysis, survival, qualitative evidence and calibrated conclusionsReview before pooling
A systematic review uses an explicit question, eligibility criteria, search, selection, appraisal and synthesis. Meta-analysis is an optional combination inside that process.
- Systematic review question and search
- Eligibility and critical appraisal
- Forest plots and study weights
- Heterogeneity and random effects
- Survival analysis and censoring
- Qualitative inference and evidence levels
Sampling determines the population represented by the data. Measurement quality limits the trust placed in every later analysis.
Statistical procedures then quantify an estimate and its uncertainty, but they cannot repair a weak question or design.
How the chapters build the chain
The opening chapters move from answerable clinical questions to experimental, quasi-experimental, observational, sampling and survey choices.
The middle chapters cover distributions, missing data, hypothesis testing, group comparisons, factorial and multivariate models, correlation and regression.
The final chapters ask whether measurements are reliable and valid, how diagnostic accuracy is calculated, and how systematic review, meta-analysis, survival analysis and qualitative research contribute different forms of evidence.
What to practise for the final
The current course materials publish a final written test made up of multiple-choice questions and short answers.
The useful revision unit is therefore a decision sequence: identify the design, outcome scale, group structure and target estimate; choose the analysis; interpret the estimate, interval and p-value; then state the design or measurement boundary. Original practice in this guide follows those published question forms without reproducing any University assessment item.
Worked example · free
Choose the test before reading the p-value
- 2Identify outcome scale, independence and group count.
- 3Choose a robust or rank-based comparison and state its target.
- 3Report descriptive distributions, an estimate and a design boundary.
Key terms
- Conceptual Model
- A diagram and argument showing how exposures, outcomes, mediators, confounders and modifiers are expected to relate.
- Internal Validity
- The extent to which a study estimate is protected from alternative explanations within the study.
- External Validity
- The extent to which an inference transports to the target people, settings and delivery conditions.
- Sampling Frame
- The operational list or source from which members can actually be selected.
- Type I Error
- Rejecting a null hypothesis that is true under the stated model.
- Type II Error
- Failing to reject a null hypothesis when the specified alternative is true.
- Interaction
- A model term showing that one effect changes across levels of another factor.
- Residual
- The observed outcome minus the value predicted by a fitted model.
- Reliability
- The consistency of measurement under specified occasions, raters or forms.
- Construct Validity
- Evidence that scores behave in ways expected from the theory of the intended construct.
- Sensitivity
- The proportion of reference-positive cases that the index test identifies as positive.
- Specificity
- The proportion of reference-negative cases that the index test identifies as negative.
RS517 FAQ
How is this course assessed?
The current course introduction allocates 20% to four online quizzes, 40% to an oral research-proposal presentation and 40% to a final written test. The timetable and presentation guide publish the current presentation and final dates. Check Canvas for live submission settings and any operational update.
What is known about the final written test?
The current course introduction says the final written test mixes multiple-choice questions and short answers. The reviewed materials do not publish a duration, section allocation or permitted-materials rule, so revision should focus on the decision methods and question forms actually taught rather than a guessed paper structure.
How should I choose a statistical test?
Start with the research question and design, then identify outcome scale, independence or pairing, and the number of groups, factors or outcomes. Distributional assumptions refine the resulting branch. A software menu or p-value should never select the test after the result is visible.
What does a non-significant p-value mean?
It means the observed result did not provide enough evidence to reject the specified null at the chosen threshold under the model. It does not prove equivalence, absence or that the null is true. Read the effect estimate and confidence interval to see which clinically important effects remain compatible with the data.
What makes the research proposal coherent?
A coherent proposal carries one line from the knowledge gap to a specific question, conceptual model, design, sample, measurement and analysis. The presentation guide also expects feasibility, timeline, justified budget, potential impact and references. Every method choice should answer a question established earlier in the presentation.
How are reliability and validity different?
Reliability concerns consistency under specified conditions, while validity concerns whether the intended interpretation and use of a score are supported. Reliability is necessary but insufficient: a measure can reproduce the same biased value. Choose reliability and validity evidence from the decision the measure must support.
How should missing data be handled?
First describe who is missing, when and why. Methods based on observed predictors can help when missingness is explained by recorded information, but they cannot verify that assumption. Report the model, compare with a simple analysis and use sensitivity analysis when unseen outcomes may drive loss.
What should I do when the presentation panel challenges a method?
Name the trade-off before defending the choice. Explain which bias or feasibility problem the method controls, what it costs, and how the analysis manages the remaining limitation. A strong response links the challenge back to the question, design and intended inference instead of reciting a definition.
Questions students ask in each RS517 chapter
- 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. Chapter 01 · Research Questions, Models and Ethics
- What is the difference between random sampling and random allocation?
- Random sampling controls selection from a population frame and supports representation of that frame. Random allocation assigns enrolled participants to study conditions and supports internal group comparability. A trial may allocate well but recruit narrowly, while a survey may sample broadly without supporting a treatment-effect claim. Chapter 02 · Experimental Designs and Validity
- Does a large sample remove sampling bias?
- A large sample can reduce random sampling error around the estimate produced by its recruitment route. It does not restore target members missing from the frame, correct systematic non-response or repair measurement error. Precision can therefore increase around an estimate that still represents the wrong group. Chapter 03 · Sampling and Survey Design
- When should the median replace the mean?
- Use the median when the typical ordered value is the target and a long tail or extreme values make the mean unrepresentative. Keep the distribution visible and report an interquartile range. The mean may still matter for totals, costs or models, so the decision follows the question. Chapter 04 · Descriptive Statistics and Missing Data
- What does p less than 0.05 mean?
- Under the null model and analysis assumptions, a result at least as incompatible with that null as the observed result would occur less than five per cent of the time. It does not mean the null has less than five per cent probability or that the effect is important. Chapter 05 · Hypothesis Tests, Errors and Power
- What decides between an independent and paired test?
- The data-generating structure decides. Use a paired route when observations are repeated on the same person or deliberately matched into meaningful pairs. Similar group sizes or baseline values do not create pairing. Treating repeated observations as independent gives the wrong standard error. Chapter 06 · Choosing Tests for Group Comparisons
- Why should the interaction be read first?
- An interaction shows that the effect of one factor changes across another. An averaged main effect can then conceal clinically important differences or even reverse direction across settings. Read the interaction estimate and simple effects before writing a general treatment sentence. Chapter 07 · Factorial and Multivariate Analysis
- Does strong correlation mean two measures agree?
- No. Two measures can rank people almost perfectly while one is consistently higher. Agreement requires examination of paired differences, systematic bias and spread across the measurement range. Correlation alone does not establish that the measures can be used interchangeably. Chapter 08 · Correlation and Regression
- Can a measure be reliable but invalid?
- Yes. A device can produce nearly the same biased reading each time, or a coherent item set can measure the wrong construct consistently. Reliability limits possible validity but does not establish content, criterion, construct or decision validity. Chapter 09 · Reliability, Validity and Diagnostic Accuracy
- Is meta-analysis always better than narrative synthesis?
- No. Pooling is useful when study questions, outcomes and effect measures are sufficiently compatible. A statistical model cannot make incomparable studies equivalent. When heterogeneity reflects important clinical differences, structured narrative synthesis or separate pools may be more informative. Chapter 10 · Evidence Synthesis and Research Inference
How to study for the exam
Revise by drawing the research chain rather than memorising test names. For each topic, write one clinical question, the design that could answer it, the unit of observation, outcome scale, main estimate, statistical route and strongest limitation. Then answer one multiple-choice item by rejecting each distractor for a precise reason and one short-answer item in four sentences: design, method, result and boundary.
Rework the same scenario across chapters so that changing the design forces a different test and changing the measurement quality changes the conclusion. For the proposal, rehearse a two-minute defence of sampling, measurement and analysis choices. Keep dates and live submission settings on Canvas separate from conceptual revision.
The final written test is published as a mix of MCQs and short answers, so fast classification and concise interpretation are more valuable than long notes that cannot be converted into a decision.
Your AI Statistics tutor for RS517
Stuck on a hard RS517 question? Sia is AskSia’s AI Statistics tutor — ask any RS517 Research Methods & Data Analysis question and get a clear, step-by-step explanation grounded in how the course is actually taught and assessed. Read this whole study guide free, then take your hardest questions to Sia.