Visual Exploration of Multifaceted Mixed Data

One possible workflow using our DimLift (lifting dimensions) approach

Visual analytics methodologies aim to attenuate the ever-growing complexity of data by combining computing with visualization techniques. Comprehensive data points capturing numerous facets of a patient cohort, a storm system, or a DNA sample allow researchers to develop new insights when paired with interaction and feedback methods from visual analysis tools. However, visual analysis of large, highly heterogeneous datasets is a considerable challenge. Most prime candidates for visual analysis are comprised of a variety of quantitative and qualitative data elements. Visual analysis often relies on the use of various summary statistics that can be used to explore and reduce the dimensionality of the dataset, but the majority of these measures are applicable only for quantitative data. Take, for example, a typical medical cohort – there are numerous quantitative measures in these data, such as height and weight, but there are also just as many, if not more, categorical measures, such as gender, education, or disease class. Visual analysis of this dataset using statistics for quantitative data without equal weighting for the qualitative data may overlook a number of interesting and clinically relevant patterns and relationships.

In this research collaboration between the University of Bergen Visualization Group in the Department of Informatics and Otto von Guericke University Magdeburg MedDigit Group in the Clinic of Neurology, we aim to extend existing visual analysis tools and methodologies for exploration of all possible data types and correlations.

Laura Garrison
Laura Garrison
Associate Professor in Visualization

My research investigates processes and assumptions designers make when crafting visualizations of complex data, and their impact on audience engagement and behavior.