Jeffrey K. Bye (California State University, Dominguez Hills), Ji Y. Son (California State University, Los Angeles)
Abstract
Null hypothesis significance testing (NHST) remains a central component of many introductory statistics courses. However, decades of research and classroom experience suggest that NHST is conceptually difficult for students, in part because it asks them to start with a hypothetical null model rather than from the data they actually observe. This workshop explores a data-first approach to teaching inference in which bootstrapped sampling distributions are treated as the primary inferential object, rather than the null hypothesis. Using student-friendly R packages and Jupyter notebooks, workshop participants will generate bootstrapped sampling distributions from real data. We will discuss how this approach can better support student understanding of estimation and confidence intervals, and subsequently, NHST. Participants will design their own bootstrapping lesson using a dataset relevant to their own students, including moments of discovery for students’ conceptual understanding. No prior experience with R is necessary for participation in this workshop.