This project was largely carried out during my Master’s Thesis at the Media Computing Group under the supervision of Prof. Chat Wacharamanotham. The project was motivated by our belief that many novice HCI researchers (e.g., grad students, early-stage PhD students) have a hard time figuring out how to choose the correct statistical procedure.
The idea behind Statsplorer is simple: Automate the selection of statistical tests. Users can upload their datasets and specify the hypothesis by selecting the measurement and factor(s). Statsplorer picks an appropriate visualization based on the characteristics of the selected variables, checks the assumptions of statistical tests, selects an appropriate statistical test, computes the results of this test, and visualizes the results. Additionally, Statsplorer also produces a statistical report of the conducted statistical test, which can be copy-pasted by the user into a report or publication.
I built the Statsplorer web application, and was involved in the research paper writing process. Additionally, I tested Statsplorer in a preliminary study with eight users, which was helpful in identifying areas of improvement and future work.
For more details (incl. demo video and our full paper), see the project page.