Project brief and cases

The group project that runs through all four parts: deliverables, milestones, six selectable cases with a minimum requirement each, and rules for data and tools.
Author

R. Heller

One project accompanies the whole course. Each group develops the same case from an evidence card in W2 to a defended recommendation in W8. The project is assessed with the rubric and counts 40% in runs of the course that are graded (proposal; the course team announces the conditions before the course starts).

What you hand in

Document Length Content
Decision sheet one page People affected, the decision and its conditions, regulatory questions, data flow, threat and control, responsibility and the next review date. Template: decision sheet.
Model passport one page Purpose, population, data unit, comparison, validation, uncertainty and limits. Template and worked example: model passport.
R script as needed A commented script with a start command (for example Rscript r/project.R) that runs in a fresh R session and reproduces every number in the passport.

The scope allows one baseline and one meaningful counter-check. For a research case, a reasoned proposal for a further experiment, or against one, is a complete result.

Milestones

After Milestone Feedback
W2 Case card (W1) and evidence card (W2) Peer feedback in unit 08
W4 First model passport with a counter-check and a decision on permissible exports Peer review in unit 16
W6 Reproducible interim report: another group reproduces one number from a fresh R session Course team, unit 24
W7 Regulatory note, data flow and control matrix for the decision sheet Role-based feedback in the tabletop exercise
W8 Defence, change request, revision note Assessment with the rubric

Six cases to choose from

Each case builds on one of the labs, so that every group starts from working code and synthetic data. Groups may propose their own case if it meets the same minimum requirement and uses only synthetic or explicitly released teaching data.

Case Starting lab Decision to support Minimum requirement
1 · Fracture alert Lab W3 Should the alert be piloted in the emergency department, and at which threshold? Baseline and one alternative on fixed test identifiers; errors at one threshold by subgroup with denominators and intervals.
2 · Hyperspectral tissue characterisation Lab W4 Is the spectral classifier ready for a study in a new species or in people? Animal-wise split with zero shared identifiers; an illumination counter-check; the limit of generalisation stated.
3 · Gait measurement from video Lab W4, gait station Can the video method replace the reference measurement for a rehabilitation decision? Agreement per person with a pre-specified tolerance; a decision on keeping or deleting raw video.
4 · Spinal cord injury biomarker Lab W5 Is the biomarker trajectory a reason for a new study or a further animal experiment? Mixed model with time unit, number of people and interaction; one sensitivity analysis for dropout.
5 · Cell typing in tissue images Lab W4, tissue station and Lab W5, cluster station Does a proposed cell subtype justify a change in how samples are reported? A threshold or scaling counter-check; results summarised per person; a check for rare feature combinations before any export.
6 · Operating capacity in a mass casualty incident Lab W6, capacity station Is a third operating theatre on standby worth its cost for the scenario studied? Two scenarios on identical pre-generated events; waiting time with its spread over replications; who waits longest.

Rules for data and tools

  • Use only synthetic data or data that are explicitly released for teaching. No patient data, no unpublished project results, no real hospitals or products as the subject of the case.
  • Label every number as synthetic, published, assumption or open.
  • If you use an AI tool (for code, text or literature), document tool, version, date, purpose and what you checked. You are responsible for every sentence and every line of code you hand in.
  • Every claim about a source needs the passage it rests on (see the evidence card in W2).

Further reading