Students in a biology curriculum at the University of Mons learn to use R, RStudio, git and GitHub in their data science courses. The flipped classroom approach is used since 2018; students discover by themselves new concepts using online course material (https://wp.sciviews.org) at home. Then they apply the new concepts in the classroom on biological datasets in individual or group projects hosted on GitHub (https://github.com/BioDataScience-Course).
{learnr} tutorials are used for a smooth transition between the theory at home and practice in the classroom. Students self-assess their conceptual understanding thanks to these tutorials with or without guided feedback using {gradethis}. {learnitdown} (https://www.sciviews.org/learnitdown/) is used to manage students identity and to record their activity in a MongoDB database.
Students in the first course are more perseverant when guided feedback is available, translating in a significant increase in the number of attempts per question. In the second course, their perseverance was already high without the guided feedback, which exhibits thus a more limited effect.
Perception of the workload in the tutorials is quantified using a Raw Task load index (NASA-LTX questionnaire). This index could be used as one of the indicators to check the improvement of the tutorials in the future.
data science class, flipped classroom, learnr, gradethis, learnitdown
Number of regular students, tutorials, questions and recorded events during the last two years are presented in the next table.
| Course | Academic year | Users | Learnr | Questions | Events |
|---|---|---|---|---|---|
| BioDataScience I | 2019-2020 | 44 | 22 | 280 | 57003 |
| BioDataScience I | 2020-2021 | 50 | 25 | 239 | 103389 |
| BioDataScience II | 2019-2020 | 31 | 8 | 93 | 17730 |
| BioDataScience II | 2020-2021 | 37 | 13 | 146 | 37875 |
| BioDataScience III | 2020-2021 | 26 | 7 | 44 | 8621 |
The learnr tutorials analysed in the present study are available in GitHub:
}}
The comparison with and without guided feedback does not show significant differences in number of attempts per exercise except for the Biodatascience I course (Wilcoxon independent test, W = 33, p-value = 2e-16). Guided feedback induces a behaviour change: students gave up sooner without them.
The NASA-LTX indicator is composed of six questions on a Likert scale to quantify the perceived workload to complete a tutorial (Hart and Staveland 1988). The questions concern mental load, physical load, time pressure, expected success, effort required, and frustration experienced during the accomplishment of the task. The average value for the six questions constitutes a Raw Task Load indeX (RTLX) (Byers, Bittner, and Hill 1989).
{{}}
Despite the increased difficulty of the exercises from one course to the other, RTLX indicator for Biodatascience III is significantly lower than for Biodatascience I (Tukey HSD, p-value = 0.023). We hypothesise that students are more used to the {learnr} tutorials as well as, R coding required to solve the exercises. Consequently, global workload is perceived as lighter. These RTLX measurements are meant to be a reference point for the following years: any future improvements in the tutorials should lead to similar or lower RTLX.
The combination of {learnr}, {gradethis} and {learnitdown} provides a powerful tool to self-assess learning in data science. Students first exposed to {learnr} tutorials also benefit from additional guided feedback with {gradethis}. We propose to use the RTLX index as a mean to assess perceived difficulties and work required to complete these tutorials.
The data will be opened shortly.