I almost skipped this decision entirely — turns out that would have been a mistake.
Why this decision matters
Moving on from my previous post on tool selection, working through the material on research paradigms and method selection pushed me to a more fundamental question: what kind of research is this actually, and what does that mean for how I analyse my findings? It would have been easy to treat method selection as a formality, pick "quantitative" because I'm working with a numerical dataset and move on. Working through the readings on qualitative, quantitative, and mixed-methods approaches convinced me that's too shallow a justification, and that the method decision has real downstream consequences for how defensible my final results will be.
Working through the decision
My project uses the Diabetes 130-US Hospitals dataset — a secondary, pre-existing quantitative dataset of 101,766 clinical admission records (Strack et al., 2014), including demographic fields like race and weight, clinical measures, and a 30-day readmission outcome.
Using the checklist-based reasoning from Urban and Van Eeden-Moorefield (2017), the decision toward a quantitative-dominant approach was fairly clear: my research question concerns measurable outcomes (imputation accuracy, downstream classification performance), not lived experience or narrative meaning that would call for qualitative data.
What surprised me was how Creswell and Plano Clark's (2018) concept of method dominance still applied even to a project I'd already decided was quantitative. There's a small interpretive element embedded in an otherwise numerical pipeline: judging why a field like race might be missing — administrative inconsistency versus a genuine data collection gap — draws on clinical and contextual reasoning that isn't purely statistical, even though the output of that judgment feeds directly into a quantitative decision (which imputation method is valid to apply). I hadn't expected a "quantitative" project to have any qualitative-adjacent reasoning in it at all, so this was a genuinely useful correction to an assumption I'd made too quickly.
What this means for my analysis plan
Greenfield and Greener's (2016) treatment of elementary statistics gave me the practical layer underneath this decision. Rather than assuming a single test will do the job, I'm planning a layered analysis:
- Descriptive statistics and distribution checks — establishing skewness and whether fields are close to normal.
- Comparative significance testing to evaluate imputation quality — chi-square for nominal categorical fields like race, and t-tests or their non-parametric equivalents for continuous fields, depending on whether normality holds.
- Classification-model performance metrics — testing whether the imputation method chosen actually matters for predicting 30-day readmission.
Given that several of my demographic fields are unlikely to be normally distributed, I'm expecting to lean more on non-parametric tests than a default t-test-first approach would assume.
Takeaway
Method selection isn't a one-off decision made at the start and then forgotten — it's something I'll need to keep justifying at each stage, from data gathering through to final significance testing.
Being upfront about the parts of it that aren't purely mechanical — like the missingness-mechanism judgment — will make the eventual write-up more honest and more defensible than pretending the whole pipeline is objectively statistical throughout.
References
Creswell, J.W. and Plano Clark, V.L. (2018) Designing and Conducting Mixed Methods Research. Third Edition. Los Angeles: SAGE.
Greenfield, T. and Greener, S. (2016) Research Methods for Postgraduates. Newark: John Wiley & Sons, Incorporated. Available at: http://ebookcentral.proquest.com/lib/sunderland/detail.action?docID=4644084 (Accessed: 19 September 2026).
Strack, B. et al. (2014) 'Impact of HbA1c Measurement on Hospital Readmission Rates: Analysis of 70,000 Clinical Database Patient Records', BioMed Research International, 2014, p. 781670. Available at: https://doi.org/10.1155/2014/781670.
Urban, J.B. and Van Eeden-Moorefield, B.M. (2017) 'Chapter 4: Choosing Whether to Use a Qualitative, Quantitative, or Mixed-Methods Approach', Designing and Proposing Your Research Project. Washington, DC: American Psychological Association.
Top comments (0)