Prof. Dr. Larry AdamsAcademic, Author & Researcher
Research Methods — Level 7 · Chapter 14 of 20

13. Data Analysis, Quality, Ethics, and Research Integrity

13.1 Analyse in relation to objectives

Analysis should be planned before data collection where possible. Each objective should have a corresponding analytic strategy. Qualitative data may be coded and developed into themes; quantitative data may be summarised and modelled; mixed-methods data require explicit integration. Researchers should not choose methods solely because software makes them available. Software assists organisation and computation but cannot replace a defensible research design or substantive interpretation.

13.2 Causal claims and confounding

Causal inference requires more than observing that two variables move together. Confounding occurs when another factor is related to both the explanatory variable and outcome. Reverse causality is also possible: engagement may influence perceived supervisor support, as well as supervisor support influencing engagement. Longitudinal, experimental, quasi-experimental, or carefully justified causal designs may improve inference, but each has assumptions. Reports should use terms such as “associated with” when causal evidence is insufficient.

13.3 Research ethics

Core ethical responsibilities include respect for persons, informed consent, minimising harm, fair recruitment, privacy, confidentiality, and responsible use of data. Participants should understand the purpose, procedures, foreseeable risks, voluntary nature of participation, data use, retention, and contact routes for questions. Special safeguards may be needed where participants depend on the researcher or institution. Research involving employees, students, children, vulnerable groups, or sensitive personal data requires particular care and appropriate review.

13.4 Privacy, digital data, and AI tools

Digital research can involve platform logs, automated transcription, online surveys, public posts, and AI-assisted analysis. Public availability does not automatically mean that data can be used without ethical consideration. Researchers should check platform terms, consent expectations, re-identification risks, and applicable data-protection rules. If AI tools are used, disclose their role according to institutional policy, protect confidential information, verify outputs against source data, and retain human responsibility for interpretation. Do not upload identifiable or restricted data to unapproved systems.

13.5 Research integrity and transparent reporting

Integrity requires accurate recording, honest analysis, appropriate attribution, disclosure of conflicts, and reporting of results that do not support expectations. Fabrication invents data; falsification manipulates data or results; plagiarism presents others’ work without proper acknowledgement. Selective reporting and undisclosed analytical changes can also mislead. Keep a secure audit trail of instruments, codebooks, exclusions, transformations, and analytic decisions. Negative or inconclusive findings can still be valuable when methods are sound.

13.6 Limitations and transferability

A limitation is not an apology; it is a guide to the boundaries of interpretation. State what the design can establish, what it cannot establish, which groups or settings may be underrepresented, and how measurement or access may affect the results. Recommendations should reflect these boundaries. A small, context-specific study may provide useful insight without supporting broad statistical generalisation; a large survey may offer broad patterns while missing the meaning and mechanisms behind them.