With Eirini Koutoumanou and Angie Wade (University College London)
Statistics teaching in a classroom of people of varying professional and educational background is challenging – more so when they are asked to visualise the same dataset and decide on the most appropriate statistical analysis that will answer a specified research question. Data generation exercises can ease this process by removing any dependency of prior understanding of the data, so leading to greater educational gains for all students. Such exercises can be easy to organise, straight forward to run and very effective in their pedagogical outcome. We will present a particular exercise that we have developed at the Centre for Applied Statistics Courses in UCL, London, UK, which involves flipping beermats and assessing the relationship between the ability to do this and height. Students perform several tasks, fill in questionnaires and the data is analysed by the course tutors and presented to the group on the final afternoon of the course. At this point, there is discussion about the design of this study, which has deliberately incorporated flaws to be highlighted, and the subsequent analyses and interpretation. This particular exercise, which we have used for many years, generates discussion about random selection, confounding, graphical presentations, competing summary statistics and inferences based on limitation of the design and analyses. Regardless of student background and prior expertise, the whole group is engaged and the exercise has proven an effective means of pulling together the formal teaching given throughout the course. We will present details about the exercise, with specific instructions of how it is performed, example outputs, student quotes, and teaching tips for other practitioners and its potential for expansion to include additional features.