Epidemiological Exploratory Data Analysis Tasks in a Geovisualization Environment

Open Access
Gruver, Adrienne Brook
Graduate Program:
Master of Science
Document Type:
Master Thesis
Date of Defense:
Committee Members:
  • Cynthia Ann Brewer, Thesis Advisor
  • exploratory spatial data analysis
  • epidemiology
  • geovisualization
  • tasks
In recent years the process of visually interacting with data has become increasingly important as a method for understanding the rich amounts of data collected. The geography of health and disease is a domain that has great potential to benefit from such interactive visualization. Geographical information systems (GIS) have already become a substantial visual tool to assist medical geographers, public health planners, epidemiologists, and others working with spatial distribution of health related data. Geovisualization builds on the utility of traditional GIS through a design meant to elicit data exploration from the user. Users can observe relationships not only across space but also examine relationships between multiple variables and over time. Geovisualization has proven to be useful for exploratory data analysis in epidemiology and public health, but it has yet to be adopted by epidemiologists and public health practitioners as a widely accepted method to explore or analyze epidemiological data. Facilitating adoption requires fitting the tool to the user and their tasks. This research is part of that larger goal to fit geovisualization tools to the needs of epidemiological analysts. Through in-depth case studies with epidemiological data researchers using a geovisualization application for data exploration and analysis, I address 1) what tasks these experts engage in while working in a geovisualization tool, 2) how they approach accomplishing their tasks, and 3) what tools, features or statistics are necessary to support their data exploration. Using task models (Knapp 1995), I organize and present the tasks of these epidemiological analysts, as well as the physical actions they took toward achieving the tasks, mental actions during the process of doing the tasks, and the visual operators utilized to achieve understanding to complete the task. Analyzing these task models for contiguity between their components, characteristics among the models and between the two participants, and what types of issues prevented the participants from moving forward in their tasks, led to the identification of five broad classes of functions necessary within a geovisualization application in order to enable completion of user tasks: 1) filters and selections, 2) making comparisons, 3) seeking statistical feedback, 4) analyzing more than two variables at a time and 5) exporting information (visually or as a table). I also established that there are several factors that influence how a user approaches accomplishing a task. Those factors include 1) the users’ visualization goals and the flexibility they have to address their goals via access to data and changing data, 2) what part of the research process they are in with the data they are exploring, 3) their experience and training with GIS and spatial methods, and 4) their usual methods of inquiry outside ESTAT. The in-depth case studies presented in this thesis provide examples of the utility of geovisualization for health-related research, and contribute evidence toward ways geovisualization can improve to meet epidemiological users’ needs.