A Pedagogical Framework for Developing Computational Social Scientists in Educational Research

• In Event: Exploring Data Science Across the Curriculum and Across Grade Levels

Abstract

During the 2018-19 school year, a series of three pilot courses were run at the University of Oregon introducing PhD students in educational research programs to data science and computational approaches to educational research. These courses operated in a cohort model, with essentially all students enrolled in the first course taking all three courses. The first course introduced students to the R programming language, git/GitHub, and the principles of reproducible research, with an emphasis on R Markdown. The final project was a group project housed on GitHub, with each member making commits and pulling/pushing to the repository to develop an APA manuscript written entirely in R Markdown, with analysis code weaved with text, making all tables, figures, and in-text references to results dynamic to the code. The second course emphasized the communication of data through a variety of mediums, with a specific emphasis on data visualization, and working with relational data. The final project included a website or online dashboard displaying the fundamental findings of a study for more than one audience, including lay audiences. Finally, the third course focused on functional programming, writing custom functions to extend the functionality of R, and developing interactive web-based data applications. The final project included either an interactive data application deployed on the web or an online tutorial, describing more efficient programming techniques.

This presentation will cover lessons learned from the delivery of these three courses, described above, including areas of successes and areas for improvement. Permission has been obtained from students to share samples of their work, including the progression of both their code written and the products produced across the three courses. Sample text from course evaluations related to the content will also be shared. We share this information within a pedagogical framework for training graduate students on computational methods in social science research, with a specific emphasis on educational researchers. Key takeaways include (a) the need to focus on programming skills separate from statistics or domain-specific courses, including direct, explicit instruction of code; (b) starting with the “low hanging fruit” of skills most directly related to students’ applied work, including data visualization; (c) placing greater emphasis on collaborative workflows with git/GitHub; (d) balancing package workflows to ease usability while not overly burdening users or their scripts with excess dependencies; and (e) placing an explicit emphasis on reproducible and open workflows, enabling the verification of not only scientific findings, but also procedures. In sum, we discuss the development of these courses, share students work and evidence of their growth, data from students’ perceptions of the courses, and our developed pedagogical framework emerging from these experiences. We view this framework as a starting point for future research on best practices in teaching t to educational researchers.

Author

• Daniel John Anderson, University of Oregon