Resources for researchers and students.

A research question is the engine of a study: it determines your design, your sample and your analysis. Start with a broad area, read enough to find a gap or a disagreement in the literature, then narrow until the question is specific enough to be answered with data you can realistically collect.

A useful checklist (FINER)

• Feasible: adequate participants, time, skills and budget.
• Interesting: to you, your field and your readers.
• Novel: confirms, refutes or extends existing knowledge.
• Ethical: can be done without unacceptable risk to participants.
• Relevant: matters to practice, policy or further research.

Specify the parts

Clinical and health questions often use PICO: Population, Intervention (or exposure), Comparison and Outcome. Even outside health, naming the population, the key concept or exposure, the comparison if any, and the outcome makes a vague question concrete.

Test it: could two researchers read your question and agree what data would answer it? If not, it needs sharpening.

The two are closely related and often confused. A research question states what you want to find out. Objectives state what you will do to find it out.

Example (illustrative)

Question: What proportion of adolescents in a district report symptoms of depression, and which factors are associated with them?

General objective: To assess depressive symptoms among adolescents in the district.

Specific objectives: (1) To estimate the prevalence of depressive symptoms using a validated scale; (2) to examine the association between depressive symptoms and selected socio-demographic factors.

Tips

• Begin objectives with measurable verbs such as estimate, compare, assess or identify, and avoid vague verbs such as understand or explore unless the design is qualitative.
• Every objective should map to a data item and an analysis step.
• Keep the number manageable: a few well-defined objectives are better than many loosely defined ones.

Design follows the question. Begin by asking what kind of answer you need.

• How common is something? A cross-sectional (prevalence) study.
• What factors are associated with an outcome, measured at one time? Cross-sectional analytic study.
• Does an exposure lead to an outcome over time? A cohort study.
• What exposures are linked to a rare outcome? A case-control study.
• Does an intervention work? A randomised or quasi-experimental design where feasible and ethical.
• What are people's experiences or reasons? Qualitative methods such as interviews or focus groups.
• What does the existing evidence show? A systematic review, with or without meta-analysis.

Then weigh feasibility, ethics, time, cost and the risk of bias. An honest, well-executed simple design is better than an ambitious design that cannot be completed.

A good review is not a list of summaries. It is an argument about the state of knowledge, built from evidence.

• Clear purpose: state the question the review is answering.
• Documented search: record databases, search terms, dates and inclusion criteria so another person could repeat it.
• Synthesis: organise by theme, debate or method rather than source by source, and show how studies agree, differ and why.
• Critical appraisal: consider study quality, sample, measurement and risk of bias, not only the findings.
• A gap that leads to your study: end by showing what is unknown and how your question addresses it.

Keep notes on every source as you read, and cite only what you have actually read.

• Mismatch between objectives and design: objectives that require comparison or causation paired with a purely descriptive design.
• Unjustified sample size: a number with no calculation, assumptions or source. Confusing the sampling method with the sample size is common.
• Undefined variables: no operational definitions, cut-offs or measurement scales.
• Unvalidated or unpiloted tools: questionnaires written from scratch with no pilot or reliability check.
• No analysis plan: tests chosen after data collection rather than planned against each objective.
• Ethics treated as a formality: missing consent procedure, confidentiality arrangements or approval details.
• Unacknowledged limitations: bias, confounding and generalisability not discussed.

A methodology review before data collection is the cheapest point at which to fix these.

Fit first

• Read the aims and scope, and look at recent issues for similar topics, designs and article types.
• Check that your article type, word limit and reporting requirements are accepted.
• Consider the readership you want to reach.

Then quality and practicalities

• Review the peer-review process, editorial board and stated timelines.
• Verify indexing claims in the relevant database itself rather than trusting a logo on the journal's website.
• Understand open-access options and any article-processing charges before submitting.

Spotting warning signs

Unsolicited emails promising very fast publication, vague scope, no named editors, unclear fees or false indexing claims are red flags. The Think. Check. Submit. checklist, the Directory of Open Access Journals (DOAJ) and the Committee on Publication Ethics (COPE) are useful references.

• Start with thanks and a short summary of the main changes.
• Respond point by point. Copy each comment, then give your reply and say where in the manuscript the change was made (page and line).
• Accept what improves the paper. Most comments, even blunt ones, point to something unclear.
• Disagree politely, with evidence. If you decline a request, explain why and cite support.
• Do not ignore a comment. If you cannot do an extra analysis, say so and explain.
• Submit a marked-up version and a clean version as the journal requests.

Keep the tone professional; the reviewers are volunteers and the editor is reading how you respond.

Plagiarism is presenting someone else's words, ideas or data as your own. It includes copying without attribution, close paraphrasing that follows the original's structure, reusing your own published work without disclosure, and citing sources you have not read.

Related breaches

• Fabrication: inventing data or references.
• Falsification: changing or omitting data or results.
• Contract cheating and impersonation: submitting work prepared by someone else as your own.
• Authorship misrepresentation: adding or omitting authors contrary to the criteria set by your field. Many journals follow the ICMJE criteria.

Good practice

• Take careful notes that separate your words from quotations.
• Paraphrase by understanding the idea and writing it afresh, then cite it.
• Treat a similarity report as a prompt for judgement, not a verdict.
• Check your institution's or journal's policy on the use and disclosure of AI tools.

Both are citation styles; they differ in how sources appear in the text and in the reference list.

APA (author-date)

• In-text: the author and year, for example (Surname, 2022).
• Reference list: alphabetical by author surname.
• Common in psychology, education and many social sciences.

Vancouver (numeric)

• In-text: numbers in order of first citation, such as [1] or superscripts depending on the journal.
• Reference list: numbered in order of citation.
• Common in medicine and the health sciences.

Always follow the exact guidelines of your university or target journal. Details such as the number of authors listed, punctuation and abbreviations of journal titles vary, and a reference manager helps only if its style file matches the required version.