The spotlight on reproducible research has spurred an increased focus on teaching reproducibility in quantitative courses, particularly statistics and data science. While there is broad agreement on the importance of reproducibility, the strategies for implementing these methods into the curriculum varies. To adequately prepare majors for their careers, reproducibility training offered solely in upper-level courses focuses more on advanced techniques and software. However, reproducibility is a vital component of other students’ curriculum and knowledge, and those students may only take a few quantitative courses, such as statistics or data science. To reach these students, reproducibility should be incorporated into our entry-level courses. Given the wide variety in entry-level curriculum and software, this means that the reproducibility techniques must be accessible and applicable to all students and, to the extent possible, be software independent. In this talk, I suggest core reproducibility strategies that are accessible to a broad student base and are software independent. These strategies will allow students to build a foundation that they can carry into their careers, regardless of what methods, tools, and software they use. Finally, I describe in further detail some alternatives to more advanced reproducibility tools such as Git. The alternatives showcase how simple tools can be used to adequately meet reproducibility criteria.
Nicholas Bussberg is an Assistant Professor of Statistics at Elon University. He received his Master’s in Environmental Science and Ph.D. in Statistical Studies from Indiana University Bloomington, where he also taught courses in introductory statistics and reproducibility. His research falls broadly into two areas: environmental statistics and spatial statistics. In environmental statistics, Nicholas uses a statistical lens to improve methodology and analysis in areas such as tree-ring analysis and marine ecosystems. His current work in spatial and spatio-temporal statistics is creating new, more flexible methods for making predictions on a sphere.