Group project — choose a question and defend an analysis

Choose a scientific question, justify the analysis and propose a follow-up experiment.

Our group project connects experimental design, simulation, analysis and communication. Work in your existing group of three. Choose one manageable primary question and target quantity, then develop an analysis as the relevant methods are introduced.

Choose a scientific context

The four cases contain source-paper links, unchanged synthetic teaching observations and factual design metadata. Their results are not findings from the papers. The starter imports and inspects data; your group chooses the question, comparison, distribution, dependence structure and model.

Case	Recorded outcomes and design	Your decision
Calf feeding and growth	Repeated body weights; individually allocated strategies within shared pens	Choose a biologically useful growth question and target.
Broiler feed efficiency	Pen feed/gain measurements; four diets within blocks	Choose a justified planned comparison or diet summary.
Pig feeding and diarrhoea	Seven-day disease counts from allocated pens within batches	Define a question and risk/comparison convention.
Horse shelter use	Group-day shelter duration and weather/insect observations	Choose an observational relationship or contrast and its interpretation.

You may propose another question and existing dataset. Show the instructor its dictionary, design and access route; keep a supplied fallback until the alternative is confirmed. Confirm the question and dataset by 4 November. A choice that cannot be supported by the observations should be narrowed after feedback.

Find out what the design supports

Collect a few useful sources on biological context, outcome definitions/units, allocation or observational sampling, independent units, grouping, time and practical constraints. For a proposed follow-up, investigate a useful magnitude of change and plausible variation. Record uncertainty when information is missing; a convenient numerical assumption is not automatically evidence.

Read the linked paper’s abstract and relevant methods first. Distinguish original study details from the invented teaching design. You need not read every section or collect new animal observations.

Submit an analysis plan before fitting

In teaching Week 10, submit the analysis-plan.qmd in your case folder before fitting the project model. It asks for your chosen question/target, design diagram and units, outcome support, proposed sampleable distribution, mean/dependence structure, planned checks, one justified alternative and information gaps. Explain every proposed term and assumption. Keep feedback and dated revisions with your final work.

Class examples and documentation provide syntax; instructor troubleshooting can help implement the approach you justify. You are not expected to invent advanced methods or use the most elaborate possible model. If computing is unavailable, after the plan is submitted the instructor can provide outputs for your declared analysis; acknowledge that support and interpret/check the results yourself. If earlier worked case outputs have been accessed, note this in the plan so an independent focus can be agreed.

Proposed educational outputs

1. A one-page scoping brief during week 42, revised after feedback.
2. An analysis plan submitted in Week 10 before fitting, with subsequent decisions documented.
3. A short reproducible report, approximately 3–5 pages including the main figure. Justify the analysis, show executed evidence from checks and one alternative, estimate the target with uncertainty, and explain design limits.
4. A follow-up experiment proposal. Choose independent allocations and measurements, a useful prospective target, a few feasible sizes and two justified assumption sets. Build a modest simulation using generic class syntax; state omitted features, unresolved fits and Monte Carlo uncertainty. Do not calculate observed power from the realised project estimate.
5. An 8-minute presentation followed by 5 minutes of questions, explaining your choices and what the evidence supports.

Final assessment requirements and deadlines will be confirmed separately. The presentation date remains to be announced; these suggestions do not establish official grading criteria.

Milestones

Date	Bring or produce
7 October, Unit 6	Provisional context and question.
12–18 October, calendar week 42	Revision/scoping brief, at most 2½ hours per student; no fitting.
21 October, Unit 7	Question/design feedback and feasible data route.
28 October, Unit 8	Audit allocation/sampling, grouping and metadata.
4 November, Unit 9	Confirm question/data or use the supplied fallback.
11 November, Unit 10	Submit the analysis plan before fitting; revise after feedback.
18 November morning, Unit 11	Implement/check your justified analysis and alternative.
18 November afternoon, Unit 12	Refine the estimate, figure and interpretation.
25 November, Unit 13	Exchange reproducible reports, review follow-up designs and rehearse.
2 December, Unit 14	Present, discuss and respond to questions.

Rotate driver, model translator and design auditor roles. Every group member should be able to explain the question, independent unit, target estimate, chosen model and one consequential limitation. Use protected workshops and proportionate technical help rather than adding model-fitting work to the autumn break.