How to choose an Appropriate Method for Research?

Choosing an appropriate research method means selecting the approach, design, sampling strategy, data-collection procedure, and analysis that can answer your research question with credible evidence. Start with the question—not a preferred tool—and then assess the required evidence, intended claim, ethics, feasibility, disciplinary standards, and limitations before finalizing the method.

Introduction

A research method should not be chosen simply because it is familiar, inexpensive, popular, or easy to use. A questionnaire is not automatically appropriate for every student project, interviews are not suitable for every question about human behaviour, and combining two methods does not automatically create a strong mixed-methods study.

The correct choice depends on what the study needs to establish. A researcher trying to estimate how common a behaviour is requires different evidence from one trying to understand how people experience that behaviour. A study testing whether an intervention causes an improvement requires a different design from one describing an existing association.

This guide explains how to choose an appropriate method for research by connecting the research problem, questions, intended claims, design, sampling, data collection, instruments, analysis, ethics, and practical constraints. It also provides comparison tables, examples, a method-fit scorecard, a justification template, and a final checklist.

Key Takeaways

• The research question should guide the method, not the other way around.
• Methodology, design, method, instrument, sampling, and analysis are related but distinct decisions.
• Qualitative research explores meaning and context; quantitative research measures patterns, differences, relationships, or effects; mixed methods integrates both.
• A method is appropriate only when the resulting data can support the intended conclusion.
• Ethics, access, measurement quality, time, skills, and resources can rule out an otherwise attractive design.
• Method selection should include the analysis and reporting plan before data collection begins.

What Is an Appropriate Research Method?

An appropriate research method is one that produces evidence capable of answering the research question while remaining ethically acceptable, methodologically defensible, feasible, and consistent with the study’s theoretical and disciplinary context.

“Appropriate” does not necessarily mean perfect. Real research is conducted under limits involving time, funding, participant access, equipment, skills, law, culture, and ethics. The aim is to choose the strongest defensible method within those constraints and to state honestly what the resulting evidence can and cannot show.

Methodology, Design, Method, Instrument, Sampling, and Analysis

These terms should not be used interchangeably.

Element	Meaning	Example
Methodology	The overall logic and justification guiding how knowledge will be produced	Quantitative, qualitative, mixed methods, participatory, design science
Research design	The structure used to answer the question and control or understand alternative explanations	Randomized experiment, cross-sectional survey, longitudinal cohort, case study
Sampling strategy	The procedure used to select participants, cases, documents, events, or observations	Simple random, stratified, purposive, theoretical, criterion sampling
Data-collection method	The procedure used to obtain evidence	Interview, observation, survey, experiment, document collection
Instrument	The specific tool used during data collection	Questionnaire, interview guide, observation schedule, test, sensor
Analysis method	The procedure used to organize and interpret evidence	Regression, thematic analysis, content analysis, survival analysis

For example, “I will use a questionnaire” is not a complete methodology. The researcher must still explain the design, population, sampling, measures, administration procedure, data-quality controls, and statistical analysis.

The Core Rule: Let the Research Question Drive the Method

The most important selection principle is simple:

Choose the method that can produce the evidence required to answer the research question.

Begin by identifying the central verb or intellectual task in the question.

• Explore suggests open-ended investigation of a poorly understood issue.
• Describe asks what exists, how often it occurs, or what its characteristics are.
• Compare examines differences between groups, conditions, places, or periods.
• Relate investigates associations among variables.
• Explain examines why or how a process occurs.
• Test an effect requires evidence about whether an intervention changes an outcome.
• Predict estimates an unknown or future outcome from available information.
• Evaluate judges the merit, effectiveness, implementation, or consequences of a policy or program.
• Understand experience examines meanings, perceptions, identities, or lived experience.
• Develop theory builds an explanation from systematically analysed evidence.
• Create and assess involves designing and evaluating an artefact, system, model, or intervention.
• Synthesize integrates findings from an existing body of research.

The wording alone does not determine the design. A question containing “effect,” for instance, does not support a causal conclusion unless the design can address selection effects, confounding, time order, and competing explanations.

Quantitative, Qualitative, or Mixed Methods?

Quantitative research

Quantitative research is appropriate when the study needs numerical evidence about frequency, magnitude, distribution, differences, relationships, prediction, or causal effects.

Common designs include:

• Experiments and randomized trials
• Quasi-experiments
• Cross-sectional surveys
• Longitudinal or cohort studies
• Case-control studies
• Structured observation
• Secondary-data analysis
• Measurement and validation studies
• Computational modelling

Its strengths include standardized measurement, numerical comparison, statistical estimation, and the possibility of generalization when design and sampling support it. Its limitations include potential loss of context, dependence on measurement quality, and vulnerability to confounding or model misspecification.

Qualitative research

Qualitative research is appropriate when the study needs to understand meaning, experience, context, interaction, identity, practice, culture, or process in depth.

Common qualitative designs include:

• Qualitative descriptive studies
• Phenomenology
• Ethnography
• Grounded theory
• Narrative inquiry
• Case study
• Discourse analysis
• Participatory and action research
• Qualitative content analysis

Its strengths include depth, contextual sensitivity, flexibility, and attention to perspectives that may not fit predefined response categories. Its limitations may include intensive data collection and analysis, researcher influence, limited statistical generalization, and the need for strong reflexivity and transparent interpretation.

Mixed-methods research

Mixed methods is appropriate when one type of evidence is insufficient and the study will deliberately integrate quantitative and qualitative components.

Integration may occur through:

• Connecting: results from one phase determine participants or questions for another.
• Building: one dataset helps develop an instrument or intervention for the next phase.
• Merging: quantitative and qualitative results are compared or combined.
• Embedding: one method is included within a larger design.
• Explaining: interviews help explain unexpected statistical findings.
• Exploring: qualitative work identifies concepts that are later measured quantitatively.

Mixed methods is not justified merely because a researcher plans to conduct a survey and several interviews. The proposal must state why both components are needed, which component has priority, when each will occur, and how the findings will be integrated (Creswell et al., 2011).

Approach	Best suited to	Typical evidence	Main limitation
Quantitative	Measurement, comparison, association, prediction, effect estimation	Numerical variables and structured records	May inadequately represent meaning or context
Qualitative	Experience, meaning, process, interaction, culture	Interviews, observations, documents, images, field notes	Does not usually estimate population prevalence
Mixed methods	Complex questions requiring breadth and depth	Integrated numerical and qualitative evidence	Requires more time, expertise, and an explicit integration plan

Match the Question to the Research Design

Research purpose or question	Potential design	Common methods	Possible analysis
How common is X in population Y?	Cross-sectional or prevalence study	Probability survey, administrative records	Proportions, confidence intervals, weighted estimates
Is X associated with Y?	Correlational, cross-sectional, or longitudinal design	Scales, records, structured observation	Correlation, regression, multilevel or longitudinal models
Does intervention X cause change in Y?	Randomized experiment or credible quasi-experiment	Intervention, pre/post measures, comparison group	Effect estimates, regression, difference-in-differences
How do people experience X?	Phenomenology or qualitative descriptive design	In-depth interviews, diaries	Phenomenological or thematic analysis
How does a process unfold?	Case study, ethnography, process study, longitudinal qualitative design	Observation, interviews, documents	Process tracing, thematic or narrative analysis
How does a group’s culture shape practice?	Ethnography	Participant observation, field notes, interviews	Iterative ethnographic interpretation
What theory explains a social process?	Grounded theory	Iterative interviews and observations	Constant comparison and theory development
How and why did an outcome occur in a bounded case?	Case study	Multiple sources of evidence	Within-case or cross-case analysis
How well is a program implemented, and what outcomes occur?	Process, outcome, or mixed-method evaluation	Records, surveys, interviews, observations	Outcome analysis plus implementation analysis
What is already known about X?	Systematic, scoping, rapid, or narrative review	Database searching and study selection	Narrative synthesis, meta-analysis, evidence mapping
Can a new system or artefact solve a defined problem?	Design science, engineering design, usability study	Prototyping, experiments, user testing	Performance metrics and qualitative feedback
How is a concept represented in texts or media?	Content, discourse, visual, or document-based research	Documents, posts, images, archives	Content, discourse, thematic, or multimodal analysis

This table offers starting points, not automatic prescriptions. The final selection must account for the field, theory, population, measurement, available comparison groups, setting, ethics, and analytic assumptions.

How to Choose an Appropriate Research Method in 12 Steps

1. Define the Research Problem

State the gap, contradiction, practical difficulty, or theoretical uncertainty that motivates the study.

A broad topic such as “social media and students” is not yet a research problem. A clearer problem might be:

Existing studies report an association between intensive social-media use and student anxiety, but little is known about how first-year students interpret the role of online comparison during the transition to university.

This problem points toward experience, interpretation, and context rather than merely estimating a numerical association.

2. Write Focused Research Questions and Objectives

Each question should be:

• Clear
• Researchable
• Sufficiently focused
• Ethically answerable
• Compatible with available evidence
• Relevant to the stated problem

Avoid choosing a method before writing the question. Otherwise, the question may be distorted to fit a convenient tool.

3. Specify the Intended Claim

Ask what the final conclusion is expected to say.

Possible claims include:

• A phenomenon exists.
• A characteristic is common or uncommon.
• Two variables are associated.
• One group differs from another.
• An intervention caused an effect.
• A model predicts an outcome.
• Participants interpret an experience in particular ways.
• A process operates through certain mechanisms.
• A program is acceptable, feasible, or effective.
• Existing evidence supports or does not support a conclusion.

The stronger the intended claim, the stronger the design requirements. A causal claim normally requires more than a one-time survey. A claim about population prevalence normally requires more than a small convenience sample.

4. Review Previous Research and Disciplinary Standards

Examine recent, high-quality studies that addressed similar questions.

Record:

• Designs used
• Populations and settings
• Measures or interview approaches
• Sampling procedures
• Analytic methods
• Limitations
• Reporting guidelines
• Unresolved methodological debates

Precedent is informative but not decisive. Repeating a common method is not justified when that method has known weaknesses or does not suit the new context.

5. Define the Unit of Analysis, Population, and Context

Determine exactly what will be studied.

The unit may be:

• An individual
• A household
• A classroom
• An organization
• A country
• A policy
• An event
• A document
• An online community
• A software system
• A published study

Confusion about the unit of analysis can produce invalid conclusions. For example, relationships observed at country level cannot automatically be attributed to individuals.

Also define:

• Target population
• Inclusion and exclusion criteria
• Setting
• Relevant period
• Geographic boundaries
• Online or offline context

6. Choose the Methodological Approach

Decide whether the study is primarily quantitative, qualitative, mixed-methods, conceptual, computational, participatory, design-based, or evidence-synthetic.

This decision should be consistent with:

• The research question
• The type of evidence required
• The theoretical framework
• Assumptions about what can be known
• The intended form of explanation
• Disciplinary expectations

A pragmatic mixed-methods approach may be suitable when the problem requires several forms of evidence. An interpretive qualitative methodology may be more coherent when the study concerns meaning and lived experience. A quantitative approach may be necessary when estimating an effect, distribution, or relationship.

7. Select a Research Design

The design determines how observations will be organized and what inferences can reasonably be made.

Ask:

• Will variables be manipulated?
• Is random assignment possible and ethical?
• Is a comparison group available?
• Is time order important?
• Will data be collected once or repeatedly?
• Is the study focused on a bounded case?
• Is immersion in a setting necessary?
• Is theory being tested or developed?
• Is an existing evidence base being synthesized?
• Is the study evaluating an artefact or intervention?

Do not label a study “experimental” merely because it includes a practical activity. A true experiment requires manipulation and an appropriate allocation or comparison structure.

8. Select Data Sources and Collection Methods

Identify the evidence that would most directly answer each question.

Potential sources include:

• Participants’ self-reports
• Behavioural observations
• Administrative records
• Sensor or laboratory measurements
• Documents and archives
• Digital trace data
• Images, audio, or video
• Existing datasets
• Published studies
• Software logs
• Physical artefacts

Then choose the collection method.

For example:

• Use observation to study what people do in an accessible setting.
• Use interviews to investigate experiences and interpretations.
• Use validated scales to measure defined constructs.
• Use experiments to estimate effects under controlled conditions.
• Use records when retrospective or population-level information already exists.
• Use documents or digital traces when representations, policies, communication, or behaviour recorded by a system are the object of study.

Consider whether self-report is likely to be affected by memory, social desirability, literacy, language, or access.

9. Choose the Sampling Strategy

Sampling should follow the question and design.

Quantitative studies may use:

• Simple random sampling
• Systematic sampling
• Stratified sampling
• Cluster or multistage sampling
• Census data
• Carefully justified non-probability sampling

Qualitative studies may use:

• Purposive sampling
• Criterion sampling
• Maximum-variation sampling
• Theoretical sampling
• Snowball sampling
• Typical, critical, or deviant case selection

A large biased sample is not necessarily more useful than a smaller, well-targeted sample. Likewise, a qualitative sample should not be defended merely by stating that qualitative research uses few participants. The researcher should explain why the selected cases provide relevant, varied, or theoretically informative evidence.

10. Plan the Analysis Before Collecting Data

A data-collection method is unsuitable when its output cannot be analysed in a way that answers the question.

For quantitative studies, define:

• Outcome and predictor variables
• Measurement level
• Primary comparison or parameter
• Statistical model
• Assumptions
• Treatment of missing data
• Potential confounders
• Subgroup analyses
• Sample-size or precision requirements

For qualitative studies, define:

• Analytic orientation
• Unit of coding or interpretation
• Role of theory
• Coding procedures
• Reflexivity
• Negative or deviant-case analysis
• Audit trail
• Team-coding or interpretive procedures where appropriate

For mixed methods, identify:

• Timing
• Priority
• Point of integration
• Procedure for handling divergent findings
• Intended combined inference

Planning the analysis in advance also helps reveal whether the proposed instrument collects the necessary variables or depth of material.

11. Evaluate Ethics, Law, Culture, and Accessibility

A method is not appropriate if it creates unjustified risk or excludes people unnecessarily.

Consider:

• Informed consent
• Privacy and confidentiality
• Data minimization
• Anonymization or pseudonymization
• Vulnerable or dependent participants
• Psychological, social, physical, economic, or legal risks
• Cultural and linguistic appropriateness
• Accessibility for people with disabilities
• Compensation and undue influence
• Community permission or governance
• Data retention and future reuse
• International data transfer
• Researcher safety
• Potential group harm or stigmatization

Human-participant research should reflect respect for persons, beneficence, and justice, alongside applicable institutional and legal requirements (National Commission, 1979).

Ethical planning should occur during method selection, not after the study has been designed.

12. Test Feasibility, Pilot the Plan, and Write the Justification

Assess whether the design can be completed with the available:

• Time
• Funding
• Participant access
• Data access
• Equipment
• Software
• Language ability
• Statistical or qualitative expertise
• Supervisory support
• Research team
• Ethics-review timeline

Feasibility does not mean choosing an unsuitable method because it is easier. Better responses may include narrowing the question, seeking collaboration, obtaining training, using a suitable existing dataset, or acknowledging a more limited objective.

Pilot the procedure where appropriate. A pilot can test recruitment, comprehension, timing, data capture, instrument performance, interview prompts, technical systems, and analysis workflow. Record what changes were made and why.

Research Question-to-Method Decision Guide

Use a survey when:

• Standardized responses are needed from many units.
• The constructs can be measured validly through self-report.
• The aim concerns distribution, comparison, or association.
• The population can be sampled and reached appropriately.

Do not use a survey solely because it is inexpensive. Avoid claiming that a convenience survey represents a wider population without evidence.

Use interviews when:

• The study needs detailed accounts, explanations, experiences, or interpretations.
• Participants can discuss the topic safely.
• The interviewer can manage probing, rapport, power, and confidentiality.
• The analysis will examine meaning rather than convert all answers into superficial counts.

Use focus groups when:

• Interaction among participants is analytically useful.
• The study concerns shared norms, disagreement, collective language, or group responses.
• The topic is suitable for discussion in front of others.

Focus groups may be unsuitable for highly sensitive disclosures or when power differences prevent open participation.

Use observation when:

• Behaviour, interaction, routines, or environmental context are central.
• Self-reports may not capture what people actually do.
• The researcher can ethically access the setting.

Observation requires decisions about participation, visibility, recording, field notes, researcher influence, and consent.

Use an experiment when:

• The purpose is to estimate an intervention’s causal effect.
• Manipulation is possible.
• A credible comparison can be created.
• Allocation, contamination, adherence, and outcome measurement can be managed ethically.

Use a case study when:

• The research concerns a bounded contemporary case.
• Context is inseparable from the phenomenon.
• Multiple evidence sources can be combined.
• The aim is an intensive explanation rather than a population estimate.

Use secondary data when:

• Suitable data already exist.
• Variable definitions and data quality fit the question.
• Access, licensing, consent, and privacy permit the analysis.
• The researcher understands how and why the data were originally generated.

Use a systematic or scoping review when:

• The question concerns an existing body of literature.
• A transparent search and selection process is possible.
• The chosen review type matches the purpose.

Systematic reviews commonly answer focused effectiveness or association questions, whereas scoping reviews often map concepts, evidence types, or research gaps. Reporting should follow an appropriate guideline such as PRISMA 2020 when applicable (Page et al., 2021).

How to Select Sampling and Sample Size

Method selection and sample-size planning are related, but they are not the same decision.

Quantitative sample size

Quantitative sample size may depend on:

• Expected effect size
• Desired precision
• Outcome variability
• Statistical power
• Significance threshold
• Number of predictors
• Repeated measures
• Clustering
• Design effect
• Expected attrition
• Planned subgroup analysis
• Population size in some designs

A generic online calculator should not be used before these elements are defined.

Qualitative sample adequacy

Qualitative sample adequacy depends on the study’s purpose and information needs rather than a universal numerical rule.

Consider:

• Specificity of the participant group
• Diversity required by the question
• Complexity of the phenomenon
• Depth of each interview or observation
• Use of multiple data sources
• Analytic approach
• Theoretical needs
• Whether additional cases still provide relevant new insight

Researchers should explain the logic of sample adequacy and document how the decision was made.

Mixed-methods samples

Mixed-method studies may require different sampling logic for each component. A large survey sample and a purposively selected interview subsample can be appropriate when the connection between them is clearly explained.

Plan Data Analysis Before Collecting Data

Quantitative alignment

The analysis depends on the question, design, variable type, distribution, dependence structure, and assumptions.

Examples include:

Question	Possible analysis
What is the average or prevalence?	Descriptive statistics and confidence intervals
Are two groups different?	Mean, proportion, or distribution comparisons
Are variables related?	Correlation or regression
What predicts an outcome?	Regression or predictive modelling with validation
Does an intervention change an outcome?	Effect estimation using a design-appropriate model
How does an outcome change over time?	Longitudinal or repeated-measures analysis
How long until an event occurs?	Survival or event-history analysis

The name of a statistical test should not be selected from the wording of the question alone.

Qualitative alignment

Common approaches include:

• Thematic analysis
• Qualitative content analysis
• Framework analysis
• Grounded-theory analysis
• Interpretative phenomenological analysis
• Narrative analysis
• Discourse analysis
• Conversation analysis
• Ethnographic interpretation

These approaches are not interchangeable. Each makes different assumptions about language, experience, context, theory, and the researcher’s role.

Mixed-method integration

Plan how the components will produce a combined conclusion. A joint display, linked sampling process, explanatory follow-up phase, or integrated discussion may help make the relationship explicit.

Use a Method-Fit Scorecard

A scorecard can make trade-offs visible, although it cannot replace methodological judgment.

Rate each candidate design from 1 to 5.

Criterion	Guiding question
Question fit	Does the design directly answer the principal question?
Inference fit	Can it support the intended descriptive, relational, causal, interpretive, or evaluative claim?
Evidence quality	Can it generate valid, credible, and sufficiently detailed evidence?
Sampling fit	Can the relevant population, cases, texts, or events be selected appropriately?
Analysis fit	Can the resulting data be analysed using defensible procedures?
Ethical acceptability	Are risks, consent, privacy, and fairness manageable?
Feasibility	Can it be completed with available time, resources, access, and skills?
Disciplinary fit	Is it defensible within the field while still allowing justified innovation?
Transparency	Can procedures, decisions, data provenance, and limitations be documented?

An optional weighted score is:

Method-fit score = Σ(weight × rating) ÷ Σ(weights)

Weights should reflect the study’s priorities. Ethical acceptability should normally operate as a minimum requirement rather than something that can be compensated for by a high score elsewhere.

Examples of Choosing an Appropriate Research Method

Example 1: Measuring the prevalence of food insecurity

Question: What proportion of undergraduate students at a university experienced food insecurity during the previous 12 months?

A cross-sectional quantitative survey may be appropriate if the researcher can use a suitable measure and reach a sample that represents the intended student population. Stratification may be needed if prevalence is expected to differ by study level or campus.

Interviews alone would provide valuable experiences but would not estimate population prevalence.

Example 2: Understanding doctoral-student isolation

Question: How do international doctoral students experience and manage social isolation during their first year?

A qualitative design using purposive sampling and in-depth interviews may be appropriate. The analysis could examine recurring patterns while preserving contextual differences.

A fixed-response survey may be useful later, but it may prematurely impose categories when the relevant experiences are not yet well understood.

Example 3: Testing a teaching intervention

Question: Does retrieval-practice software improve examination performance compared with standard revision materials?

A randomized or credible quasi-experimental design is preferable because the intended claim concerns an effect. The design should specify allocation, baseline differences, adherence, outcome measurement, attrition, and contamination.

A post-intervention satisfaction survey cannot establish whether the software improved performance.

Example 4: Evaluating implementation of a public-health program

Questions: Did vaccination uptake increase, and how did staff and residents experience implementation?

A mixed-method evaluation may be appropriate. Administrative data can estimate changes in uptake, while interviews or observations can investigate access barriers, implementation processes, and unintended consequences.

The two components should be integrated rather than reported as unrelated mini-studies.

Example 5: Developing a digital research tool

Question: Can a new citation-checking system identify incomplete references accurately and fit researchers’ workflows?

A design-science or engineering approach could include iterative development, benchmark testing, error analysis, usability observation, and user interviews. Performance and usability require different but complementary forms of evidence.

Example 6: Studying online community norms

Question: How do members of an online support community negotiate acceptable forms of advice?

A digital ethnography or discourse-oriented design may be appropriate. Ethical issues include whether the space is reasonably perceived as public, whether quotations are searchable, how consent will be approached, and how members will be protected from identification.

Ethical and Practical Considerations

Participant access

Do not build a design around a population that cannot realistically be reached. Confirm permissions, recruitment routes, gatekeeper requirements, and likely response rates before committing to the method.

Measurement quality

Use instruments with evidence relevant to the target construct, language, population, and setting. A scale validated in one cultural or clinical context is not automatically valid in another.

When adapting an instrument, document translation, cognitive testing, pilot results, scoring changes, and consequences for interpretation.

Validity and credibility

In quantitative research, consider:

• Selection bias
• Measurement error
• Confounding
• Attrition
• Missing data
• Model assumptions
• External validity

In qualitative research, consider:

• Reflexivity
• Depth and variation
• Context
• Negative cases
• Transparency of interpretation
• Data adequacy
• Coherence between question, methodology, and analysis

Quality should be judged according to the logic of the chosen approach rather than by imposing one tradition’s criteria mechanically on another.

Researcher skills

Lack of current expertise is a feasibility issue, not evidence that an inappropriate method becomes suitable.

Possible solutions include:

• Methodological training
• Statistical consultation
• Qualitative supervision
• Interdisciplinary collaboration
• Narrowing the question
• Using a simpler but still appropriate design
• Conducting a pilot

Multiple feasible methods

Sometimes several methods can answer the same broad question. The researcher should compare what each method would reveal, obscure, assume, and permit them to conclude.

Method selection is therefore an argued decision, not a search for one universally correct answer.

Modern Research Tools, Artificial Intelligence, and Open Science

Digital research tools

Digital tools can support:

• Database searching and reference management
• Online surveys and electronic data capture
• Audio or video recording
• Transcription
• Qualitative data organization
• Statistical analysis
• Reproducible coding
• Version control
• Secure data storage
• Preregistration
• Collaborative project management
• Data and material sharing

Software should be chosen after the methodological requirements are known. Access to a particular program should not determine the research question.

Artificial intelligence

Generative AI may assist with limited tasks such as:

• Generating preliminary search-term alternatives
• Explaining unfamiliar methodological terminology
• Checking code syntax
• Creating simulated examples for training
• Reformatting non-sensitive materials
• Suggesting questions for human review

It should not be treated as an autonomous methodologist, source of verified references, substitute for statistical or qualitative expertise, or final judge of research quality.

Before using AI, researchers should consider:

• Institutional and journal policies
• Confidentiality and data-protection rules
• Whether participant data may be entered into the system
• Accuracy and fabrication risk
• Bias and representativeness
• Reproducibility
• Disclosure requirements
• Human verification and accountability

Sensitive, confidential, identifiable, copyrighted, or embargoed data should not be entered into an external AI service unless this is explicitly permitted and appropriately protected. UNESCO recommends a human-centred approach to generative AI in research and education (UNESCO, 2023).

Preregistration and transparent workflows

For suitable confirmatory studies, preregistration can record questions, hypotheses, outcomes, exclusions, and analyses before results are known. It does not make a weak method strong, and exploratory research remains valuable when it is reported transparently.

Open-research practices may include:

• Protocol sharing
• Preregistration
• Data-management plans
• Transparent materials
• Analytic-code sharing
• Data availability statements
• Version control
• Clear reporting of deviations
• Reusable metadata

The FAIR principles encourage research objects to be findable, accessible, interoperable, and reusable while recognizing that ethical, legal, and contractual restrictions may limit open access (Wilkinson et al., 2016).

Select reporting guidance during planning

Reporting guidelines can expose missing design details before data collection.

Examples include:

• CONSORT for randomized trials
• STROBE for observational studies
• PRISMA for systematic reviews
• COREQ for interview and focus-group studies
• SRQR for qualitative research
• SPIRIT for trial protocols
• Discipline-specific APA Journal Article Reporting Standards

A reporting checklist improves transparency; it does not itself guarantee that the underlying design is valid.

Common Mistakes When Choosing a Research Method

Starting with a favourite method

“I want to conduct interviews” is not a research problem. Begin with the question and required evidence.

Treating an instrument as the methodology

A questionnaire is an instrument. It does not specify the design, sampling, administration, measurement model, or analysis.

Choosing according to whether data are “numbers or words” alone

Methodological choice also depends on the intended claim, theory, context, sampling, time structure, comparison, and analysis.

Claiming causality from an observational association

Statistical significance does not convert a cross-sectional association into a causal effect.

Using mixed methods without integration

Two datasets do not create mixed-methods research unless their relationship and combined contribution are explained.

Ignoring analysis until after data collection

This may produce variables, transcripts, or observations that cannot answer the question.

Selecting convenience sampling without limiting the conclusion

Convenience samples may be useful for defined exploratory or local purposes, but broader generalization requires strong justification.

Choosing an underpowered or uninformative design because it is feasible

A study that cannot answer its question may waste participant time and other resources. Narrow the question or redesign the study.

Copying a previous study without examining context

A method appropriate in one population, institution, country, language, or period may be inappropriate elsewhere.

Confusing methodological limitations with poor execution

Every method has boundaries. Poor planning, weak measurement, undocumented decisions, or unsupported claims are not unavoidable limitations.

Allowing software or AI to make the decision

Tools implement or support parts of a method. They cannot determine the appropriate inference, ethical balance, theoretical coherence, or disciplinary contribution without accountable human judgment.

Research-Method Justification Template

Use the following structure in a proposal or methodology chapter:

This study uses a [methodological approach] and a [research design] because the research aims to [explore/describe/compare/explain/test/evaluate/develop] [phenomenon or relationship]. The design is appropriate for the question, “[research question]”, because it will generate [type of evidence] from [population, cases, documents, or setting]. Participants or cases will be selected through [sampling strategy], and data will be collected using [method and instrument]. The data will be analysed using [analysis procedure], which is suitable because [brief analytic rationale]. The main limitations are [limitations], which will be addressed through [mitigation]. Ethical issues involving [consent, privacy, risk, fairness, or data governance] will be managed by [procedure].

A strong justification explains why the method fits. It does not simply define the method or list its advantages.

Final Method-Selection Checklist

Before approving the methodology, confirm that:

• The problem and question are clearly defined.
• The intended conclusion is explicit.
• The methodology fits the theoretical and epistemological position.
• The design can support the intended inference.
• The unit of analysis is correct.
• The population or case boundaries are defined.
• Sampling fits the question and design.
• The data source can provide the required evidence.
• The instrument measures or elicits the relevant phenomenon.
• The analysis has been planned in advance.
• Important validity threats or credibility issues are addressed.
• Ethical, legal, cultural, and accessibility requirements are manageable.
• Participant and data access have been checked.
• Time, budget, software, equipment, and expertise are sufficient.
• A pilot or feasibility assessment is planned where appropriate.
• Mixed-method components have an integration plan.
• A relevant reporting guideline has been identified.
• Data-management, transparency, and disclosure procedures are documented.
• The limitations are stated honestly.
• The final claims will remain within what the evidence can support.

Conclusion

To choose an appropriate method for research, begin with the problem, question, and intended conclusion. Then select a coherent methodology, design, sample, data source, instrument, and analysis that can produce the required evidence. Test the plan for ethics, validity, credibility, feasibility, transparency, and disciplinary fit.

The strongest method is not the most complicated or fashionable one. It is the method that answers the question directly, manages important threats to interpretation, protects participants and data, and supports conclusions no stronger than the evidence allows.