Effects of Data Complexity and Map Abstraction on the Perception of Patterns in Infectious Disease Animations
Open Access
- Author:
- McCabe, Craig Andrew
- Graduate Program:
- Geography
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- None
- Committee Members:
- Dr Alan M Mac Eachren, Thesis Advisor/Co-Advisor
Alan Maceachren, Thesis Advisor/Co-Advisor - Keywords:
- cartogram
choropleth
averaging
aggregation
time-series
cartography
animated maps
spatial epidemiology
infectious disease
Niger - Abstract:
- Geographic map animations have become an increasingly popular method for exploring spatio-temporal datasets. While some have questioned the effectiveness of animations compared to the static alternatives, little research has been done on our ability to retain and recall data presented in animated maps. This study uses a dataset of weekly measles infections in Niger from 1995-2004 in a controlled experiment to test the ability of participants to complete a series of map-reading tasks using choropleth and schematic map representations. Infectious disease data is often imperfectly sampled and inherently noisy. As a result, an important question for map animation designers is whether to represent the raw data directly or to transform the data to dampen the impact of noise. To address this question, the experiment compares animations using the raw weekly data to transformed data sets using temporal averaging and aggregation. These approaches were used to determine what effects data complexity and spatial abstraction have on our ability to perceive spatio-temporal patterns in map animations. Experiment participants (N = 96) were recruited from undergraduate geography classes at Penn State University, then divided into groups according to three data smoothing approaches: raw (weekly) data, 5-week moving-window average, and 2-week aggregation. Participants viewed a series of animations of yearly measles epidemics using three different map representations, then provided quantitative and qualitative assessments of the spatial and temporal characteristics of the infection patterns. The results showed that overall, map representation had a larger impact on task performance than data complexity, due primarily to the perceptual salience of large, but low-population districts in Niger. It was also found that temporal aggregation, an approach used by infectious disease researchers, did not result in any significant interpretation errors overall, despite expectations to the contrary. In some tasks, such as determining when the peak of a yearly epidemic occurred, aggregation proved to be most effective, enabling participants to identify the correct peak week 50% more often, compared to those who used raw data. The results also support the idea that animated maps are best used for simple map-reading tasks or gaining a broad overview of a dataset, and less effective for making quantitative comparisons.