A conference talk at OZCOTS 2021 given by Damjan Vukcevic.
Joint work with Margarita Moreno-Betancur, John Carlin, Sue Finch, Ian Gordon and Lyle Gurrin.
Statistical teaching often focuses on models and techniques, with much less time devoted to emphasising the primacy of the real-world questions that these are meant to answer. One consequence is what we call the 'true model myth': the belief that statistics is about finding the 'best' model or technique to apply to the data.
To help students avoid this, we propose explicitly teaching the idea of a 'statistical investigation', which would be rather like a scientific investigation. This starts with well-posed questions, to be investigated via potentially several different analyses (analogous to scientific experiment), with the results drawn together to form the final conclusions.
The examples we provide should reinforce this. When we teach new techniques, we often keep things simple and show only a single analysis. We can remind students of the larger story by replacing or supplementing these with examples where multiple techniques are used. We present several such examples and exercises that would fit naturally within existing curricula.
Although this change might seem to complicate teaching, we expect that it will actually lead to greater student satisfaction since students will develop skills for "scaffolding" problems in the context of real-world questions.
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0
International License.
