|
Abstract:
|
Exploratory Data Analysis (EDA) and report writing are fundamentals skills every statistician has to master. Yet, the thought processes that lead from EDA to reasoning with data are still largely unknown. The ongoing debate in the sciences of how statistics can be misused (p-hacking, etc.) focuses on erroneous conclusions, but does not take a descriptive look on how people actually work with their data. Using our novel e-learning platform, we have built a data exploration tool that enables students to analyze data and write reports in the same environment. By collecting all interactions with the platform (the statistics students look at, what plots they generate, and the text chunks they write - along with their associated timestamps), the process of data reasoning can be investigated in detail. For example, we study how students perform model selection (e.g., which predictors they pick for a linear model and in what order) and how this fits into the narrative of their reports, or what features of the data students miss. Finally, combining the collected data from our platform with student grades, the teacher's evaluation process becomes a focus of analysis as well.
|
