Reading Groups
Are you interested in reading more about this year's eCOTS: What's Next? Moving Forward - Even better, would you like to discuss it with fellow Statistics educators? Your chance has arrived!
- R Pedagogy Reading Group
When:
Wednesday 29th May at 15:00 BST / 10:00 EDT / 07:00 PDT - Jenny Terry (RoSE Director; Brighton, UK)
Tuesday 4th June at 15:00 BST / 10:00 EDT / 07:00 PDT - Nicola Rennie (RoSE Statistics Software SIG Lead; Lancaster, UK)
Monday 17th June at 15:00 BST / 10:00 EDT / 07:00 PDT - Alyssa Counsell (RoSE Statistics Software SIG Lead; Toronto, Canada)
Monday 24th June at 21:00 BST / 16:00 EDT / 13:00 PDT - Anna Fergusson (RoSE Statistics Software SIG Lead; Auckland, New Zealand)
Thursday 4th July at 16:00 BST / 11:00 EDT / 08:00 PDT - Hilary Watt (RoSE Statistics Pedagogy SIG Lead; London, UK)
Facilitator: Jenny Terry (Jenny.Terry@sussex.ac.uk)- Statistics is a required course for undergraduate students of most social and physical sciences in universities throughout the world, including a requirement to learn how to conduct statistical analyses using appropriate software. To this end, departments are increasingly considering teaching the software, R. Not only is R free to use, but unlike point-and-click software such as the popular IBM SPSS Statistics, it is readily extendable to cutting-edge, advanced quantitative methods, and generates the kind of reproducible, easily shared analyses that are becoming the norm due to the proliferation of Open Science practices across the sciences. However, R is a programming language meaning users must learn to code - a skill most undergraduates will be yet to develop. There is, therefore, a steep learning curve, making many statistics educators hesitant to introduce it. The fact that R is challenging for novice coders is not, by itself, a reason not to teach it, but we must accept that there is also a learning curve for instructors too, for the pedagogies used to teach point-and-click software are not necessarily the most effective when teaching novices to code. R pedagogy is an emergent field and the scholarship that is out there is hard to find. To begin to address this, the RoSE Network’s Statistics Software SIG would like to take this opportunity to share and discuss some recently published R pedagogy papers, towards building a directory of work in this field and inspiring further research.
- AI in Statistics Education
When: May 29th, June 5th, June 19th, June 26th and July 03rd at 11am (New York)
Facilitator: Florian Berens (florian.berens@uni-tuebingen.de)- Artificial intelligence (AI) is already changing our lives and will continue to do so. Since the publication of ChatGPT at the latest, the public has also become aware of this development. To not only observe but actively design this development in the field of education, educational research is needed. In some fields of learning research, there already exist interesting projects and results of such research. In the field of statistics education, however, this research is still rare. This reading group therefore brings together people who are interested in researching artificial intelligence as an instrument of statistics education. Together we will read texts on artificial intelligence in educational contexts other than statistics and discuss how the results can inspire research in statistics education. The first two sessions will be introductory to enter the world of research on AI in education. The we cover two major sub-topics, namely Intelligent Tutoring Systems and Learning Analytics. In the last session we look on the relationship of AI to affects. The RoSE Network’s AI in Statistics Education SIG would like to take this opportunity to share and discuss some recently published papers, towards building a directory of work in this field and inspiring further research.
- What do students learn from simulations in statistics?
When:May 22, 5 am Philippine Standard Time
This is our first meeting and we may proceed with the same schedule for the next 4 to 5 sessions, each of which lasts an hour or less. If you have questions about this reading group, please send Andrew Pua an email at andrewpua@outlook.com !
May 21, 2 pm PT and 5 pm EDT
May 21, 10 pm British Standard Time
Facilitators: Andrew Pua (andrewypua@outlook.com)- Simulation-based inference arose as a pragmatic response to the accessibility of technology and the frustrations attached to teaching a difficult topic like hypothesis testing. Rather than rushing through hypothesis testing at the end and leaving such a crucial topic misunderstood, it does feel straightforward to introduce hypothesis testing earlier so that there are opportunities to correct misunderstandings as early as possible. Unfortunately, the most difficult aspect of starting with a simulation-based approach to conducting a hypothesis test is to think hard about the model/data generating process under the null. In other words, we need to let students develop skills to understand how to generate artificial data under the null.
Statistical methodology papers would usually include Monte Carlo simulation experiments. But a casual search of the phrase “Monte Carlo simulation” in most statistics education research journals has surprisingly fewer hits relative to “simulation-based inference”. Moving beyond hypothesis testing, students inevitably would want to explore or would find themselves in a position where more complex procedures would be used, even without necessarily having a solid conceptual foundation. It seems imperative that students pick up skills related to simulating generative processes and understanding the behavior of different procedures encountered in statistical inference.
By reading articles on simulation-based inference, especially details of its implementation and evaluation and understanding the difficulties and misconceptions which could arise in the minds of our students, we are inevitably going to move into the territory of how to design Monte Carlo simulations.
- Simulation-based inference arose as a pragmatic response to the accessibility of technology and the frustrations attached to teaching a difficult topic like hypothesis testing. Rather than rushing through hypothesis testing at the end and leaving such a crucial topic misunderstood, it does feel straightforward to introduce hypothesis testing earlier so that there are opportunities to correct misunderstandings as early as possible. Unfortunately, the most difficult aspect of starting with a simulation-based approach to conducting a hypothesis test is to think hard about the model/data generating process under the null. In other words, we need to let students develop skills to understand how to generate artificial data under the null.