After a Good Fit, What Should We Measure?
A short note on choosing the next observation when several parameter explanations already fit the data.
Topic
Teaching & research practice
Learning, mentoring, evidence and research design.
Start with an article
Filter writing on this topic →Keep the Result That Changes the Question
A failed check can narrow a useful question. Two published numerical studies show why the original failure must remain visible.
When Solar Power Changes Fast: Why Better Intervals Are Still Not Reliable Enough
A five-station study asks whether solar prediction intervals remain trustworthy during rapid power changes and missing sensor inputs. Better calibration improves every station, yet the hardest groups still fall far below the intended coverage.
When a Numerical Method Creates Negative Matter
A stiff transfer-network benchmark compares standard explicit integrators with a linearly implicit update that preserves non-negativity and total mass.
Equation Discovery That Survives a Change of Units
A controlled symbolic-regression benchmark asks whether dimensional constraints improve exact equation recovery, shrink the search space, and prevent unit-fragile formulas.
From Open Question to Reproducible Model
A compact workflow for assumptions, model choice, computation, validation, limitations, and communication.
Go a little deeper
Research programmes
Mathematical Modeling Education
Pathways from an open-ended question to assumptions, computation, critique, and communication.
Project overviews
2026
Dimensionally Constrained Equation Discovery
Known physical laws form a controlled benchmark for unit-aware monomial symbolic regression.
2026
Occupancy-Aware Classroom Ventilation
A shared-capacity predictive-control benchmark for classroom CO₂ exposure when actual room use departs from the timetable.
2026
Positive and Conservative Time Stepping
A stiff transfer-network benchmark compares explicit methods with a positivity- and mass-preserving linearly implicit update.
2026
Coastal Defence Under Rising Tides
A teaching case combines sea-level scenarios, extreme-value storm surges, and cost optimization over a 50-year horizon.
2026
When Solar Power Changes Fast
An accepted ADMA 2026 short paper on conditional solar prediction-interval reliability. Five-station comparisons distinguish consistent calibration improvement from the absolute reliability that difficult events still lack.