The Complete Guide to Research Methodology: Designs, Methods and Examples

1. Methodology, Methods and Research Design

Methodology is the reasoning that explains how knowledge will be produced and why the selected approach is appropriate. Methods are the procedures used, such as surveys, interviews, experiments or document analysis. Research design is the overall plan that connects the question, setting, sampling, collection and analysis.

2. Research Philosophy Explained Simply

Philosophy matters when it changes what counts as evidence and what claims can be made. A thesis should not include pages of philosophy disconnected from the design.

3. Choosing a Design from the Research Question

4. The Methodology Alignment Chain

Example from healthcare: a researcher asks whether a medication reminder reduces missed doses. This requires an effect estimate, so an experimental or quasi-experimental design is more suitable than interviews alone. Interviews may explain why the reminder works, but cannot independently estimate its effect.

5. The Research Onion: Moving from Philosophy to Procedures

The research onion is useful because it forces the researcher to make decisions in the correct order. Begin with assumptions about knowledge, then select an approach to theory, methodology, strategy, time horizon and finally specific procedures.

6. Ontology and Epistemology Without Jargon

Ontology concerns what kind of reality the researcher assumes exists. Epistemology concerns how that reality can be known. These ideas matter because they shape what counts as evidence.

7. Deduction, Induction and Abduction

8. Research Question to Methodology Matrix

9. Cross-Domain Research Questions and Suitable Designs

5. Survey Research

Survey research is appropriate when a study needs standardised information from a defined population. It can estimate prevalence, compare groups or test relationships, but a cross-sectional survey normally cannot establish causal direction.

Business example

A researcher examines whether perceived AI transparency predicts employee trust across UK financial-services firms. A survey can measure both constructs and test their relationship. It cannot prove that transparency caused trust unless the design addresses temporal order and alternative explanations.

Scale reliability: Cronbach's alpha = .86
Mean transparency score = 3.42 (SD = 0.71)
Correlation with trust: r = .48, p < .001
Regression coefficient: beta = .39, 95% CI [.28, .50]

6. Experimental and Quasi-Experimental Designs

A true experiment manipulates an intervention, includes a comparison condition and uses random assignment. Quasi-experiments estimate effects without full randomisation, using techniques such as matched comparison groups, interrupted time series or difference-in-differences.

7. Cross-Sectional, Longitudinal and Correlational Designs

Cross-sectional studies observe variables at one point in time. Longitudinal studies observe change across time. Correlational designs test associations without manipulation.

8. Quantitative Sampling

Probability sampling supports statistical generalisation when every eligible unit has a known chance of selection. Non-probability sampling may be necessary but requires careful limits on inference.

• Simple random: equal selection probability.
• Stratified: sampling within important subgroups.
• Cluster: selecting groups such as schools or hospitals.
• Systematic: selecting every kth unit after a random start.
• Convenience: accessible participants; high selection-bias risk.

Multiple regression
Expected effect size f² = 0.15
Predictors = 8
Alpha = .05
Power = .80
Minimum required sample = 109

9. Measurement, Reliability and Validity

A statistically sophisticated analysis cannot rescue poor measurement. Define each construct, select or develop indicators, pilot the instrument and examine reliability and validity.

10. Designing a High-Quality Questionnaire

A questionnaire should translate clearly defined constructs into items that respondents can understand and answer consistently. Begin with a construct map, not a blank survey screen.

1. Define each construct conceptually.
2. Identify dimensions.
3. Review validated measures.
4. Draft unambiguous items.
5. Select a response scale.
6. Conduct expert review.
7. Run cognitive interviews or pilot testing.
8. Assess reliability and validity.

11. Cohort and Case-Control Designs

Healthcare example: A case-control study compares patients with and without a rare adverse drug reaction, then examines prior exposure. Odds ratios are appropriate; incidence cannot be estimated directly.

12. Quantitative Analysis Workflow

13. Cronbach's Alpha: Output and Interpretation

Reliability Statistics
Cronbach's Alpha              .842
McDonald's Omega              .858
Number of Items                  8

Item-total correlations: .41 to .69
Alpha if item deleted: .80 to .84

Example thesis wording: “The eight-item transparency scale showed good internal consistency (α = .84; ω = .86). Corrected item-total correlations ranged from .41 to .69, and deleting any item did not materially improve reliability.”

14. Correlation: Output and Interpretation

                     Transparency   Trust   Adoption
Transparency             1.00       .48***   .36***
Trust                     .48***     1.00     .55***
Adoption                  .36***     .55***   1.00
*** p < .001, N = 312

15. Multiple Regression: Output and Interpretation

Outcome: Adoption intention
R² = .42, Adjusted R² = .41, F(4,307) = 55.60, p < .001

Predictor                 B       SE      Beta      p
Transparency             .28     .05      .31     <.001
Trust                    .44     .06      .46     <.001
Age                     -.01     .00     -.08      .071
Prior AI experience      .16     .07      .10      .024

Examiner question: Were residual assumptions, multicollinearity, influential observations and theoretical ordering checked?

16. Logistic Regression

Outcome: Adoption (1 = yes)
Predictor             Odds Ratio    95% CI          p
Transparency             1.82       1.39-2.39      <.001
Training received        2.41       1.48-3.92      <.001
Years in role            0.97       0.93-1.01       .122

17. ANOVA and Group Comparisons

Outcome: Trust score
F(2, 207) = 8.64, p < .001, η² = .077

Post-hoc Tukey:
High transparency vs Low: Mean difference = .58, p < .001
High vs Medium:             .24, p = .091
Medium vs Low:              .34, p = .018

18. Chi-Square Test

Training received × Adoption
χ²(1, N = 280) = 14.92, p < .001
Cramer's V = .23

19. Exploratory and Confirmatory Factor Analysis

KMO = .89
Bartlett's test: χ²(190) = 1844.30, p < .001
Parallel analysis retained 3 factors
Primary loadings = .56 to .84
Cross-loadings < .30

χ²(167) = 286.40, p < .001
CFI = .954
TLI = .946
RMSEA = .048, 90% CI [.039, .057]
SRMR = .041

20. Mediation and Moderation

X = Transparency, M = Trust, Y = Adoption
Indirect effect = .19
Bootstrap 95% CI [.12, .28]
Direct effect = .11, p = .041

Transparency × AI experience interaction
B = .14, SE = .05, p = .006
Simple slopes:
Low experience:  B = .18, p = .031
High experience: B = .46, p < .001

21. SEM and PLS-SEM

22. Quantitative Results Reporting Template

1. Describe the sample and missing data.
2. Report data screening and assumptions.
3. Report measurement quality.
4. Present descriptive statistics.
5. Report the primary inferential model.
6. Include effect sizes and confidence intervals.
7. Explain robustness or sensitivity analyses.
8. State which hypotheses were supported.
9. Separate statistical significance from practical meaning.
10. Bound causal claims according to design.

23. Multilevel and Hierarchical Models

Multilevel models are appropriate when observations are nested, such as students within schools, patients within hospitals or repeated measurements within individuals.

Outcome: Student achievement
School-level variance = 18.40
Student-level variance = 81.60
Intraclass correlation (ICC) = .184

Fixed effects:
Teaching quality      B = .31, p < .001
Class size            B = -.12, p = .018

24. Time-Series and Interrupted Time-Series Analysis

Time-series methods analyse observations collected repeatedly over time. Interrupted time series estimates whether an intervention changes the level or trend of an outcome.

Pre-intervention monthly trend      B = -0.02, p = .441
Immediate level change              B = -1.84, p < .001
Post-intervention trend change      B = -0.09, p = .012

25. Survival Analysis

Survival analysis is used when the outcome is time until an event, such as relapse, equipment failure, employee exit or publication acceptance.

Predictor                Hazard Ratio   95% CI        p
Training received             0.68      0.52-0.89   .005
High workload                 1.47      1.16-1.86   .001

26. Missing Data and Multiple Imputation

Complete-case analysis can bias results when missingness is systematic. Researchers should describe missing-data patterns and justify the selected response.

27. Measurement Invariance

Measurement invariance tests whether a construct is measured comparably across groups or time.

Configural model: CFI = .958, RMSEA = .046
Metric model:     ΔCFI = -.003
Scalar model:     ΔCFI = -.007

28. Bayesian Analysis

Bayesian analysis combines prior information with observed data to produce posterior distributions. It is useful when uncertainty, prior evidence or small-sample estimation matters.

Posterior mean effect = 0.34
95% credible interval = [0.12, 0.56]
P(effect > 0) = .997

10. Case Study Research

Case study investigates a bounded contemporary phenomenon in its real context, often using multiple evidence sources. The case is not merely the location; it is the analytical unit being investigated.

Public-administration example

A comparative case study examines how three municipalities govern algorithmic welfare decisions. Interviews, policy documents, observation and system records allow the researcher to compare mechanisms across institutional settings.

11. Phenomenology

Phenomenology examines lived experience and how people make sense of a phenomenon. It is suitable when experience itself is central, not when the goal is to develop organisational theory or evaluate programme effectiveness.

Healthcare example: How do first-time intensive-care patients experience loss of control during mechanical ventilation?

12. Grounded Theory

Grounded theory aims to develop an explanatory theory through iterative data collection and analysis. Core features include constant comparison, theoretical sampling, memo writing and category development.

Interview excerpt: “The system suggested rejection, but I overrode it because the candidate's experience was unusual.”
Initial code: overriding automated recommendation
Focused code: professional discretion against algorithmic advice
Category: negotiated authority
Emerging proposition: accountability ownership shapes willingness to override AI

13. Ethnography and Observation

Ethnography examines culture, practice and meaning through sustained engagement. Observation can reveal discrepancies between formal procedures and actual behaviour.

Engineering example: An organisational ethnography studies how safety rules are adapted during maintenance shutdowns. Interviews alone may capture official accounts; observation reveals how teams negotiate risk under time pressure.

14. Narrative and Discourse Approaches

Narrative research examines how stories organise identity and experience. Discourse analysis examines how language constructs realities, positions actors and enables power.

15. Thematic, Content and Framework Analysis

Reflexive thematic analysis develops patterns of shared meaning through active researcher interpretation. Content analysis may focus on systematic categorisation and frequency or meaning. Framework analysis uses matrices to compare cases and themes, often in applied research.

                 Site A            Site B            Site C
Decision authority Local discretion  Central approval  Hybrid model
Data visibility     High              Moderate          High
Accountability      Individual        Organisational    Shared

Qualitative sampling

Purposive sampling selects information-rich participants or cases. Maximum-variation sampling explores diversity; criterion sampling uses defined eligibility; theoretical sampling follows emerging concepts. Sample adequacy should be justified through informational needs, not a universal number.

Trustworthiness

• Credibility: are interpretations plausible and well supported?
• Dependability: is the process transparent and logical?
• Confirmability: can readers see how interpretations arose?
• Transferability: is enough context provided for readers to judge relevance elsewhere?
• Reflexivity: does the researcher examine their role and assumptions?

16. Designing High-Quality Interviews

Interviews should be designed around the analytical purpose, not merely the topic. Good questions invite detailed accounts, examples, tensions and meaning.

Interview sequence

1. Begin with context and experience.
2. Move to concrete incidents.
3. Probe reasoning and consequences.
4. Explore contradiction and change.
5. Close with reflection and omitted issues.

17. Focus Groups

Focus groups are useful when interaction itself produces insight. They are not simply a cheaper substitute for interviews.

18. From Raw Transcript to Analytical Themes

Transcript:
“When the dashboard showed red, managers called immediately. We started changing decisions just to avoid appearing risky.”

Initial codes:
- dashboard triggers intervention
- avoiding visible risk
- changing decisions under scrutiny

Focused code:
- adapting practice to algorithmic visibility

Category:
- behavioural compliance under digital surveillance

Theme:
- visibility converts advisory analytics into informal control

19. Reflexive Thematic Analysis

Reflexive thematic analysis treats themes as developed through active interpretation. It does not assume that themes simply emerge from data.

1. Familiarise yourself with the dataset.
2. Generate meaningful codes.
3. Construct candidate themes.
4. Review themes against data and research purpose.
5. Define the central organising concept of each theme.
6. Write an analytical narrative.

20. Open, Axial and Selective Coding

21. Using NVivo, ATLAS.ti or MAXQDA Responsibly

Qualitative software stores, retrieves and organises data. It does not perform the intellectual work of interpretation.

• Create a transparent codebook where appropriate.
• Use memos to record analytic decisions.
• Retain links between codes, excerpts and emerging claims.
• Use queries to test patterns, not to replace close reading.
• Export an audit trail for reporting and examination.

22. Reflexivity in Practice

Reflexivity explains how the researcher's position, assumptions and relationships influence the research process.

23. Qualitative Findings Reporting Template

1. Explain the analytical approach.
2. Describe how codes and themes were developed.
3. Present each theme as an analytical claim.
4. Use selected evidence, including variation and contradiction.
5. Explain relationships between themes.
6. Connect themes to research questions.
7. Distinguish findings from later theoretical discussion.

24. Conversation Analysis

Conversation analysis examines how social actions are accomplished turn by turn. It requires detailed transcription and close attention to sequence, timing and repair.

Manager:  So the model says decline?
Analyst:  It says high risk.
Manager:  Right, but not decline.
Analyst:  No, that's our interpretation.

25. Documentary and Archival Analysis

Documents are not neutral containers of information. Researchers should examine provenance, purpose, audience, omissions and institutional context.

26. Choosing Among Qualitative Approaches

16. Mixed-Methods Research

Mixed methods integrates qualitative and quantitative evidence to produce an inference that neither strand could provide alone. Merely conducting a survey and interviews does not create a mixed-methods study unless integration is explicit.

17. Action Research and Participatory Research

Action research combines inquiry and change through iterative cycles of planning, action, observation and reflection. Participatory approaches involve stakeholders in defining problems and producing knowledge.

Education example: Teachers and researchers collaboratively redesign formative assessment, implement it across two cycles and analyse changes in practice and student engagement.

18. Delphi Method

Delphi uses repeated rounds of structured expert judgement, often to develop consensus, priorities or forecasts. The method requires transparent expert-selection criteria, controlled feedback and a pre-defined consensus rule.

Round 1: 42 proposed competencies
Round 2: 18 competencies reached median ≥ 4
Consensus threshold: IQR ≤ 1
Round 3: 14 competencies retained

19. Design Science Research

Design science develops and evaluates an artefact intended to solve a relevant problem. Common artefacts include models, methods, software, frameworks and decision tools.

Computer-science example: A researcher develops a privacy-preserving clinical triage system. The thesis must establish problem relevance, design requirements, artefact construction, evaluation criteria and theoretical contribution.

20. Systematic Reviews, Scoping Reviews and Meta-Analysis

A systematic review answers a focused question using transparent, reproducible searching, selection and synthesis. A scoping review maps concepts, evidence and gaps. Meta-analysis statistically combines compatible effect estimates.

21. Comparative, Longitudinal and Evaluation Designs

Comparative designs explain similarities and differences across cases. Longitudinal designs investigate change. Evaluation research judges programme implementation, outcomes or mechanisms.

Public-policy example: A comparative longitudinal study examines how two cities implemented the same homelessness policy over four years, combining administrative outcomes with interviews to explain divergent results.

22. How to Integrate Mixed-Methods Evidence

Integration is the defining feature of mixed methods. It can occur during sampling, instrument design, analysis or interpretation.

Finding                Quantitative result       Qualitative explanation
Low adoption            38% regular use          Staff feared accountability
Training effect         OR = 2.41, p < .001      Training built confidence
Site variation          Significant, p = .012    Local managers framed AI differently

23. Action Research Spiral

Each cycle should generate both practical change and research learning. Document what changed between cycles and why.

24. Designing a Rigorous Delphi Study

1. Define the problem requiring structured expert judgement.
2. Set expert inclusion criteria.
3. Design Round 1 for idea generation or initial rating.
4. Provide controlled anonymous feedback.
5. Predefine consensus and stopping rules.
6. Report attrition across rounds.
7. Distinguish consensus from validity or truth.

25. Design Science Lifecycle

26. Systematic Review and PRISMA Example

Records identified: 2,460
Duplicates removed: 410
Titles/abstracts screened: 2,050
Full texts assessed: 186
Studies excluded after full text: 132
Studies included in synthesis: 54
Studies included in meta-analysis: 31

27. Programme and Policy Evaluation Designs

28. Multiple and Embedded Case Designs

A multiple-case design compares cases to test or refine explanation. An embedded design examines several units within one broader case.

29. Meta-Analysis: Output and Interpretation

Studies included: 31
Pooled standardised mean difference = 0.42
95% CI [0.27, 0.57]
I² = 61%
τ² = 0.09

30. Realist Evaluation

Realist evaluation asks what works, for whom, in what circumstances and why. It develops context-mechanism-outcome configurations.

31. Economic Evaluation

22. Research Quality Across Methodologies

23. Ethics, Privacy and Responsible Research

Ethics includes consent, risk, confidentiality, fairness, power relationships, data minimisation, secure storage and responsible reporting. Approval is not a substitute for ethical judgement throughout the project.

AI and digital-data example

A study analysing student discussion forums must consider whether participants reasonably expected research use, whether re-identification is possible, how quoted text will be protected and whether automated analysis introduces bias.

24. Methodology Quality Scorecard

25. Methodology Chapter Template

1. Restate the research purpose and questions.
2. Explain the methodological approach and philosophical position.
3. Justify the research design against alternatives.
4. Define population, setting and sampling.
5. Describe data collection and instruments.
6. Explain analytical procedures step by step.
7. Present quality criteria and validation procedures.
8. Explain ethics, privacy and data management.
9. Discuss methodological limitations and their consequences.
10. Summarise how the design answers each question.

26. Common Examiner Criticisms

• The methodology describes procedures but does not justify them.
• The sample is convenient but conclusions are generalised broadly.
• The study claims causality from cross-sectional associations.
• The qualitative analysis lacks transparency between coding and themes.
• Mixed methods are conducted in parallel but never integrated.
• Theoretical framework and analytical strategy are disconnected.
• Validity or trustworthiness is discussed generically rather than demonstrated.
• Limitations are listed but their consequences are not analysed.

27. Frequently Asked Questions

What is the difference between methodology and method?

Methodology is the logic and justification of the research approach. Methods are the specific procedures used to collect and analyse evidence.

Is quantitative research more rigorous than qualitative research?

No. Rigour depends on whether the design is appropriate and executed transparently. Different questions require different evidence.

Is mixed methods always better?

No. It adds value only when integration answers a question that one approach cannot answer adequately.

How large should a qualitative sample be?

There is no universal number. Adequacy depends on study purpose, participant specificity, data richness, analytical approach and the claims being made.

Can cross-sectional research establish causality?

Usually not on its own because temporal order and confounding are difficult to establish.

Do I need a research philosophy section?

Include philosophy when it materially explains your evidence and claims. Avoid a detached textbook discussion.

What makes a case study rigorous?

Clear case boundaries, justified case selection, multiple evidence sources, transparent analysis and a defensible analytic generalisation.

How should I justify software?

Software supports analysis but is not the methodology. Justify the analytical procedure, then explain how software assisted it.

Can AI analyse my research data?

AI may support coding or diagnostics, but confidentiality, bias, transparency, verification and institutional policy must be addressed. The researcher remains accountable.

28. Final Methodology Checklist

Print This Guide and Checklist

Important: This guide provides general educational information. Methodological expectations vary by discipline, institution, journal and research purpose. Always follow applicable regulations, ethics approvals and supervisory guidance.

29. Sampling Planner

30. Research Ethics Checklist

31. Methodology Justification Template

This study adopts a [methodological approach] because the research question seeks to [type of answer]. A [design] was selected because it enables [required evidence or inference]. Alternative approaches such as [alternative] were considered but were less suitable because [reason]. The sample was selected using [strategy] to ensure [adequacy rationale]. Data were collected through [methods] and analysed using [procedure]. Credibility was strengthened through [quality procedures], while limitations include [specific boundaries].

32. Examiner Methodology Scorecard

33. Methodology Decision Workbook

1. Write the exact research question.
2. Underline the verb: measure, compare, explain, understand, develop, evaluate.
3. State the type of claim required.
4. List the evidence necessary for that claim.
5. Identify feasible data sources.
6. Compare at least two plausible designs.
7. Choose the design with the strongest fit, not the greatest complexity.
8. State the main threat to credibility.
9. Specify how that threat will be managed.
10. Define the limits of inference before collecting data.

34. Final Methodology Audit

A strong methodology chapter should allow an examiner to reconstruct the logic of the study and understand why the conclusions are credible within stated boundaries.

35. Methodology Comparison Atlas

36. Ready-to-Use Methodology Reporting Library

Quantitative design wording

A cross-sectional explanatory design was used to estimate associations between perceived algorithmic transparency, trust and adoption intention. This design was appropriate for testing the proposed relationships across a defined professional population, although temporal and causal claims remain limited.

Qualitative design wording

A comparative qualitative case-study design was selected because the research sought to explain how institutional context shaped the use of algorithmic recommendations. Multiple cases enabled comparison of recurring mechanisms while preserving contextual detail.

Mixed-methods wording

A sequential explanatory mixed-methods design was used. Survey analysis first identified adoption patterns and site differences; follow-up interviews then explained the organisational mechanisms underlying those results. Integration occurred during participant selection, joint display construction and final interpretation.

Sampling wording

Participants were selected purposively to maximise variation in professional role, organisational setting and experience with the system. Recruitment continued until the dataset provided sufficient conceptual depth to explain the central process rather than to satisfy a predetermined numerical target.

37. Statistical Diagnostic Snapshots

Predictor                  Tolerance     VIF
Transparency                 .62        1.61
Trust                        .57        1.75
Training                     .83        1.20
Experience                   .78        1.28

Shapiro-Wilk residual test: W = .992, p = .118
Breusch-Pagan: χ²(4) = 6.21, p = .184
Durbin-Watson = 1.96
Cook's distance maximum = .18

38. Qualitative Audit-Trail Example

39. Responding to Methodology Criticism

40. Downloadable Methodology Tools

The package includes editable starter files for planning and reviewing methodology:

• Methodology decision worksheet
• Sampling planner
• Methodology audit checklist

41. Version History

42. Reporting Standards by Study Type

These frameworks support transparent reporting but do not replace methodological judgement.

43. Reproducibility and Research Transparency

• Pre-register hypotheses and analysis plans where appropriate.
• Retain code, syntax and decision logs.
• Document data cleaning and exclusions.
• Use version control for analysis files.
• Share de-identified data or synthetic data where ethical and lawful.
• Distinguish confirmatory from exploratory analysis.
• Report deviations from the original plan.

44. Open Science Checklist