Sunday, Aug 4: 4:00 PM - 5:50 PM
1577
Topic-Contributed Paper Session
Oregon Convention Center
Room: CC-B110
Bayesian inference is a flourishing paradigm in statistics. Dramatic improvements in algorithms and methodology have led to extensive work in Bayesian research. However, Bayesian methods are not broadly used across science, technology, engineering, or mathematics (STEM). To fill this need, the NSF has sponsored a faculty development initiative to promote the teaching of Bayesian methods across all STEM disciplines—the Bayes BATS program. This session will give an overview of the BATS project from the project organizers, and report on educational innovations developed by faculty participants in the program. Session attendees will leave with concrete ideas and open-access materials for teaching Bayesian statistics in their courses, drawn from a variety of disciplines.
Applied
Yes
Main Sponsor
Section on Statistics and Data Science Education
Co Sponsors
International Society for Bayesian Analysis (ISBA)
Section on Bayesian Statistical Science
Presentations
Bayesian statistics has evolved from a small sub-field to a major area of statistics and Bayesian methods are increasingly adopted in both academia and industry. However, the exposure to Bayesian statistics of undergraduate students in many STEM fields is still very limited. I give an overview of BayesBATS, an NSF-supported program aimed at training educators across a range of US institutions to enable them to introduce Bayesian methods in their curriculum. I will describe the program's activities and report on the experience of its first cohort.
We present an activity for an introductory statistics course that illuminates the differences between Frequentist and Bayesian approaches. The activity has students analyze a dataset related to social inequality and discover interesting trends—with a surprise at the end. Different student groups will come to different conclusions, which will require them to grapple with the differences between Frequentist and Bayesian approaches. We briefly describe the intervention and report results from a pilot run of the activity.
We present a series of Inquiry Based Learning activities for teaching Bayesian statistics to undergraduate learners. We discuss our use of the POGIL (Process Oriented Guided Inquiry Learning) framework to develop and facilitate the activities and report initial feedback from pilot runs of the activities in three diverse institutional contexts: an R2 Hispanic Serving Institution, a small liberal arts college, and an R1 public land-grant university.
A Bayesian statistical framework offers a versatile approach, allowing scientists to update their beliefs as new data emerges. Integrating Bayesian concepts into the STEM curriculum prior to graduate school will equip students with practical tools, applicable in diverse fields, encouraging adaptable statistical reasoning in the generations to come. To encourage science educators to infuse their courses with Bayesian reasoning, we provide access to an Open Educational Resource (OER) website that includes lesson plans, examples from various disciplines, and project-based learning activities.
Economic statistics courses prepare students to analyze data using exclusively the frequentist method, despite the increasing use of the Bayesian approach in economics applications. Inspired by Rittle-Johnson and Star's evidence (2009, 2011, 2020) on the educational benefits of comparative learning in mathematics, this paper utilizes the structure of an existing undergraduate economic statistics course, which emphasizes the frequentist approach, to introduce the Bayesian approach in parallel. Comparisons were made using the Beta-Binomial model for proportions and the Normal-Normal model for means. Applied exercises were conducted using R Studio and the bayesrules package. The main purpose of this teaching approach is not only to familiarize students with both methodologies but also to enhance their understanding of concepts that are often misunderstood in the frequentist framework, such as p-values, confidence intervals, and the implicit assumptions used to derive the distribution of estimators. I also present reflections on the benefits and challenges I encountered during my first implementation of this teaching format.