Unit 9 — Model checking and defensible conclusions
Wednesday 4 November 2026, 12:30–16:00
Use design information and diagnostic evidence to assess whether an analysis supports its conclusion.
No new video, reading or coding is required. Bring your Week 8 annotated fit, checking list and laptop if available. The teaching slides, case worksheets and results are provided in class.
Opening recall and teaching ask whether the model respects the experiment, fits reliably and describes the observations. We use performance directly to check models, lme4 to represent shared pen differences, and marginaleffects to compare predicted diets. Four cases cover extra count variation, animals sharing pens, a curved weight relationship and an unreliable positive/negative-screen fit. Justify one action, check again and state what remains uncertain.
Breaks: 13:40–13:55 and 14:55–15:10. Finish at 16:00.
At 15:45–15:50, confirm each group’s project question, primary outcome and usable dataset. Switch to the documented supplied case if access or essential design information is missing. Give short diagnostic reports at 15:50–15:55, then answer the individual exit question and look ahead to repeated measurements at 15:55–16:00. Project launch and supplied cases
Teaching slides Workshop slides 09-01 · Diagnose the model Case 1 · Pen event counts Open the instructor-supplied Posit Cloud project for 09-model-diagnostics. Choose any of the four case pages in the in-class handout and work through more as time permits. Use the diagnostic report template to record your evidence. Report template
Allow 60–75 minutes for the consolidation slides (25 minutes), targeted reading from Touchon’s second edition (25–35 minutes), and the diagnostic report and repeated-measurement sketch (10–15 minutes).
Consolidation slides 09-02 · Diagnostic report Repeated-measurements worksheet
Bring your report, sketches of the three named animals and a question about similarity through time to Class 10.