Assessing the Usability of Statistical Software Using a Discrete Choice Experiment

The usability [2, 3] of modern statistical software is increasingly important as users with limited statistical training become more reliant on such tools to help them address complex real-world problems [1, 13]. This paper presents a case study that illustrates how a discrete choice experiment [16] may be used to assess the usability of the complex controls that are becoming more prevalent in such software. The case study focuses on an interactive plot intended for novice users of a commercially available statistical application, who need to interpret a linear model used to analyze the results of an experiment. Effective selection of model terms is critical for identifying important experimental factors and the accuracy of model predictions. The ability to easily determine whether terms should be entered into a model, or the reason for their inclusion, is essential for interpretability [4, 15] and to ensure user satisfaction of the application.