Analytical Leadership Across Student Research
This is a body of methodological work, not a claim of ownership over students’ research questions. Across projects in morphology, spinal health, obstetrics, forensic estimation and comparative anatomy, I translated questions into inspectable analyses and built infrastructure that could be reviewed, reused and handed on.
01 · Problem
What needed solving
Student research often combines a strong domain question with small, complex or irregular datasets. The difficult part is choosing a defensible analysis, exposing data-quality risks and creating outputs the researcher can understand and explain.
Make each project’s reasoning explicit while building shared methods that reduce repeated work across the research group.
02 · Constraints
The difficult parts
- Work with small, irregular datasets without overstating what a model can generalise.
- Detect observer effects and other measurement problems before testing the biological question.
- Prevent leakage when classifying or estimating a held-out specimen.
- Design shared infrastructure that remains understandable when another researcher extends it.
03 · Approach
How the work was structured
- Staged data cleaning and quality-control reports before modelling.
- Used observer-controlled tests, false-discovery-rate correction and sensitivity subsets where the design required them.
- Applied grouped cross-validation, balanced metrics and genuinely held-out classification to keep validation honest.
- Extended standard shape analysis with methods such as sliding semilandmarks, spherical harmonics, block balancing, bending energy and rule-based screening.
- Delivered interactive exploration tools, documented outputs and reusable pipelines that appeared in later project folders.
04 · Validation
What can be claimed now
The archive contains analytical code across more than fifteen student research projects and course material revised over six years. Several pipelines were reused or extended by others, and blind or held-out validation was completed where the research design called for it.
The work supported theses and collaborative outputs across several disciplines while creating shared analytical infrastructure for the group. The public case study describes methodological patterns only, not private findings.
05 · Learning
What the project clarified
- The right contribution claim is methodological: build and explain the analysis while respecting ownership of the research question.
- Observer error can change the question; it should be modelled, not treated as a cosmetic limitation.
- A handover is part of the method when the person using the workflow did not write it.