Online Program

Return to main conference page
Friday, February 16
PS2 Poster Session 2 and Refreshments Fri, Feb 16, 5:15 PM - 6:30 PM
Salons F-I

Exploratory Analyses from Different Forms of Interactive Visualizations (303634)

Michelle Dowling, Viriginia Tech (Dept of Computer Science) 
Peter Hauck, Virginia Tech (Discovery Analytics Center) 
Leanna House, Virginia Tech (Dept of Statistics) 
*Lata Kodali, Virginia Tech 
Scotland Leman, Virginia Tech (Dept of Statistics) 
Chris North, Virginia Tech (Dept of Computer Science) 

Keywords: Bayesian, Multidimensional Scaling, Visual Analytics, Education, Uncertainty

Meaningful visualizations of big data are crucial. We developed a method called Bayesian Visual Analytics (BaVA) in order for analysts of any level to explore high-dimensional data interactively (House et al. 2010). One cutting edge software created under this framework called Andromeda displays multiple linear projections based on Weighted Multidimensional Scaling (WMDS) and user interactions (Endert et al. 2011; Self et al. 2016; Kruskal 1964). The direction for which data are projected in WMDS is determined by weights assigned to each variable, and interactions include surface level, parametric, object level, or any combination of these (Leman et al. 2013). With interactivity, practitioners can discover which variables best explain groupings or how observations are similar or dissimilar in settings such as customer service or textiles. There is uncertainty in these visualizations on the variable weights, the user interactions, and the WMDS projection. We aim to quantify these uncertainties using a probabilistic approach under the BaVA framework and present findings from a case study on novice analysts using Andromeda. We provide access to this versatile and easy to use software.