# Project and follow-up design peer review, 14:35–14:55
Exchange with the next group: calf → broiler → pigs → horses → calf. Use your own dataset and design, not the relay fixture. Spend 12 minutes auditing and 8 minutes discussing two priorities.

## Trace the existing analysis
- State the data source and whether values are synthetic; published context is not the source of invented outcomes.
- Locate one outcome, main comparison, time/population, units, estimate and 95% interval.
- Draw treatment allocation or observational sampling; identify animals or groups that received separate treatment assignments, or were selected separately for observation and repeated/grouped records.
- Read the saved analysis plan and decision log. Explain why the proposed distribution used to describe the outcome and model structure fit the known question/design; identify a plausible alternative and the consequences of its assumptions.
- Trace one diagnostic finding to a decision and a scientifically justified sensitivity comparison. Different defensible models are acceptable; a copied formula or agreement with an instructor result is not evidence of justification.
- Check that code assistance and any prior access to complete case solutions are disclosed; distinguish your own choices from supplied choices.
- Check the main figure and narrative use the same which group is subtracted from which and scale. Distinguish uncertainty for a mean from variation between future animals.
- For observational shelter/weather data, preserve an association interpretation; adjustment alone does not establish causation.

## Audit the future experiment
| Feature | Proposed design | Evidence or provisional assumption | Concern / action |
|---|---|---|---|
| Primary outcome and useful effect | | | |
| Allocated unit and the number of animals or groups assigned to each treatment or selected for observation | | | |
| Randomization, blocking and shared environment | | | |
| Measurement period and missing/removal policy | | | |
| Three candidate sample sizes | | | |
| Two plausible variability/dependence assumptions | | | |
| Simulation method, repetitions and seed | | | |
| Simplifications relative to final analysis | | | |

A future temperature association and a randomized shelter improvement answer different questions. More measurements on the same animal/pen/group do not automatically create more treatment replication. A retrospective calculation using an observed favourable effect does not establish an adequate future experiment. Treat simulation as prospective screening with justified assumptions; a real animal experiment requires a fuller protocol.

Return **two concrete comments**: one consequential design/interpretation issue and one reproducibility/communication issue. Name file or figure locations, why the issue matters and a feasible action. Mark unresolved assumptions rather than inventing documentation.

## Acknowledge assistance

Acknowledge code assistance and any previously accessed worked solutions. Distinguish supplied choices from your own reasoning. Be ready to explain and critique those choices individually or discuss a fresh comparison agreed with the instructor.