09-02 · Diagnostic report and repeated-measures bridge
Allow 60–75 minutes. Spend 25 minutes with the consolidation slides, 25–35 minutes reading and 10–15 minutes finishing the report and sketch.
Consolidation slides Diagnostic report template · Repeated-measurements worksheet · Study cases
Recap the model checks — 25 minutes
Use the consolidation slides and the output on the case page you’re reporting on. Ask three questions. Does the model account for how researchers assigned the diets and which observations share conditions? Did the fitting calculation work? Does the model describe the observations?
In the growth models, a residual is observed gain minus fitted gain, in g/day. Look for patterns against fitted gain, starting weight and pen. For the event counts, the model checks compare the observations with simulated counts. Under a suitable model these residuals should spread fairly evenly from 0 to 1. Explain what each plot compares before interpreting a pattern.
Use the experiment and the measured outcome to decide what to try. Keep evidence you found reassuring and check it again alongside the concern. Read the diet estimate with its interval. For the welfare-screen case, complete separation leaves the room-adjusted diet estimate unresolved. Report the observed rates by diet and explain what the intervals describe.
The last part of the deck shows how to read an animal’s measurements across several weeks and where to look for conditions shared by animals in the same pen.
Targeted reading — 25–35 minutes
Use Touchon, Applied Statistics with R, second edition. Revisit the model-checking discussion around the worked example most relevant to your case in Chapter 7 (Generalized Linear Models) or Chapter 8 (Mixed Effects Models). If you worked on the starting-weight case, revisit interpreting the mean relationship in Chapter 6. Spend about 15–20 minutes connecting one output/plot to class evidence.
Then spend 10–15 minutes on the introductory repeated-measures material in Chapter 9, identifying the design issue before developing our later longitudinal analysis. That chapter includes repeated-measures ANOVA. Our later analysis will consider animals sharing pens and observations from the same animal at nearby times. Book contents · Chapter 9.
Write your diagnostic report
Start this report in class. After class, spend 10–15 minutes finishing the report and making the sketch below. Use the displayed output and plots from your chosen case to support each answer.
Name or group: __________ Case: __________ Date: __________
- Study and starting model. State the study question, what each row measures and in what units, how researchers assigned diets, and which observations share conditions. Give the starting model’s formula and explain what it assumes.
- Evidence and possible causes. Record what the fitting output showed. Name one concerning pattern and one reassuring feature, explain what you expected to see, and give two possible explanations for the concern. Say what information would help distinguish them.
- Action and reason. Explain your model change or reporting choice and which feature of the study justifies it. State the assumption it adds or changes.
- Recheck. Compare the fitting output and diagnostic plots. Compare the diet estimates and intervals where the models can estimate them. Say what improved and what remains unresolved. For the welfare-screen case, report observed rates with their intervals and state that the room-adjusted estimate is unresolved.
- Biological conclusion and limitation. In two sentences, say what the evidence supports for this study and where uncertainty limits the conclusion. An interval that includes zero for a difference or one for a ratio also includes other possible effects.
Sketch three animals’ repeated measurements
Use the repeated-measurements worksheet to sketch three animals’ gains. Answer its interpretation questions, then reveal the example plot and answers to check your work.
Bring your diagnostic report, sketch and one question about similarity through time to Class 10.