Looks Okay to Me: A Study of Best Practice in Data Analysis Code Review
When and Where
Speakers
Description
Education in statistical computing requires that we train students not only in programming skills and principles, but also in good data science habits. In this study, we investigate the question of how certain habits, routines, or intuitions contribute to quality analysis, in the context of identifying errors in an existing work. Volunteers from two populations—professional data scientists, and mid-degree college students—were supplied with pre-populated RMarkdown notebooks, and asked to comb through the reports’ code and discussion in search of errors. We then conducted a qualitative analysis of subject behavior during the study, based on video recordings of these sessions. Ultimately, we identified many common themes in how the subjects interacted with the code, text, and IDE during their error-checking process.
About Kelly Bodwin
Professor Bodwin’s primary areas of research are the development of open-source tools for data science education, and clustering/community detection methodology for biological and social science data. Some of my current projects include a novel data mining method for large-scale binary data, an R package for automatically generating teaching materials in R Markdown and Shiny, a cross-disciplinary study of social networks in historical political groups, and a collaborative analysis of soil experiments in local vineyards.
Kelly Bodwin is an Assistant Professor of Statistics and Data Science at California Polytechnic State University in San Luis Obispo. Prof. Bodwin’s current research interests include cross-disciplinary work in the Digital Humanities, methodologies for high-dimensional clustering, and Data Science education. Prof. Bodwin is a Certified RStudio Trainer, and many of her course materials are free and open-source.