Research guide

PhD Data Analysis: From Raw Data to Defensible Findings

Data analysis is where doctoral projects most often stall. This guide covers the decisions that determine whether your analysis survives examination.

Published 15 January 2025 · Updated 15 January 2025 · 9 min read · By the MaquishTech Venture® research consulting team

Analysis starts at design time, not after collection

The worst time to choose an analysis method is when you are staring at a completed dataset. Your analysis plan should be written before data collection: what questions the data must answer, which variables or coding scheme that requires, and which tests or techniques apply.

Every survey item, interview question and data field should map back to a research question. If it does not, it is decoration — and reviewers will notice when decoration appears in your results chapter.

Choosing a statistical method (quantitative)

  • Comparing two groups → t-test or Mann-Whitney U
  • Comparing three or more groups → ANOVA or Kruskal-Wallis
  • Relationship between two variables → correlation (Pearson/Spearman)
  • Predicting an outcome → regression (linear, logistic, multiple)
  • Testing a theoretical model → structural equation modelling (SEM)
  • Always report effect size and confidence intervals, not just p-values

Qualitative analysis: coding done properly

Qualitative analysis means systematic coding — not "reading transcripts and highlighting quotes". The standard workflow: familiarise → generate initial codes → search for themes → review themes → define and name themes → produce the analysis.

Document your coding framework: first-cycle codes, second-cycle patterns, the codebook itself. An audit trail is what converts "I read the transcripts" into defensible qualitative analysis.

Documentation that survives examination

  • Keep a raw dataset that is never edited — all cleaning happens in a documented pipeline
  • Record every exclusion decision and its reason
  • Version your analysis scripts or syntax files
  • Distinguish exploratory findings from confirmatory tests
  • Report assumptions checked (normality, homogeneity) — not just the test itself
  • Interpretation must connect findings back to the research questions and literature, not just describe them

Frequently asked questions

Consultants can advise on method selection, help you structure workflows and review your analysis — but you must understand and be able to defend every analytical decision in your viva. Work done on your behalf must comply with your institution's rules on permitted assistance.

It depends on the test and expected effect size — run a power analysis (e.g., G*Power) during design. Underpowered studies are a common examiner criticism.

Whatever your institution supports and you can defend. R and Python offer reproducible pipelines; SPSS is common in social sciences and easier to learn. The method matters more than the tool.

Working on your own research?

We provide ethical research consulting — methodology guidance, data-analysis support, literature strategy and documentation workflows. You remain the sole author of your academic work; we do not write or complete assessed work for researchers.

Request an Ethical Research Consultation Discuss Your Research Challenge

Related research guides