Unit 5 — From design to simulated data
Wednesday 30 September, 12:30–16:00
What data could our experimental design generate? We will use a blocked 16-pen experiment to explore Gaussian, binary, count and positive continuous responses, working in four groups of three.
No preparation video or new reading is required. Bring your laptop and access to the course Posit Cloud project. We will recap the GLM ideas and R randomisation code needed for the activity during class.
We will start with a opening recall of experimental units and possible response values, then work through the essential GLM components and blocked randomisation.
Teaching slides Workshop slides
Each group will record its assumptions, complete a supplied R script, generate two possible datasets under the same model, and interpret or diagnose the planned GLM. We will finish with an individual exit question.
05-01 · Simulation activity Open the instructor-supplied Posit Cloud project for 05-data-simulation. Its scripts, data and reference files are already provided; open your assigned script there.
The activity page includes the four group scripts as individual downloads. Use your assigned script in Posit Cloud. Breaks are at 13:40–13:55 and 14:55–15:10.
The longer GLM deck explains the models used in class, including links, response variation, effect interpretation and difficulties with sparse binary data. It also introduces the repeated simulations we will use for power in Unit 6.
GLMs and modelling slides 05-02 · Consolidation and preparation for power
Allow about 60 minutes for the explanation and practice. Work through the deck and follow-up worksheet, then bring your corrected script, one assumption to investigate, and a proposed detection rule to the next class. A video link will be added when the recording is available.
No additional textbook purchase or reading is required for this follow-up.