Griffith

Dr. Daniel A. Griffith

Two researchers in UT Dallas’ School of Economics, Political and Policy Sciences will explore uncertainty in spatial data with a grant awarded by the National Institutes of Health.

The project focuses on documenting, visualizing and utilizing data error and uncertainty information in spatial analysis, said Dr. Daniel A. Griffith, Ashbel Smith Professor of geospatial information sciences (GIS) and the principal investigator. Dr. Yongwan Chun, an assistant professor of GIS, will serve as co-investigator on the project.

Uncertainty refers to how much the data — and, subsequently, the results of a statistical analysis — can be trusted. There are many sources of uncertainty; the researchers will investigate how the potential sources impact public health data analyses.

“When data undergo spatial aggregation, corruptions introduced through the process are not documented,” Griffith said. “Data users are not aware of the magnitude of error in and uncertainty accompanying a given data set.

“Ignoring error in spatial data is detrimental to the formulation of effective policies and the making of sound decisions.”

The collaborative grant with George Mason University totals approximately $304,000 for one year, and the expected project total for three years is more than $860,000.

The researchers also will investigate how to effectively visualize uncertainty.

yongwan

Dr. Yongwan Chun

Griffith said existing mapping tools fail to sufficiently include high-quality information. This is particularly true of health data, such as geocoded cases in databases like the Florida Cancer Registry, with which the researchers will work.

Data users often ignore data error, treating spatial data and associated maps as error-free, Griffith said. Some analyses, such as geographic cluster detections — like a crime hotspot or spatial distribution of cancer rates — are performed without considering the quality of data.

The researchers aim to:

Enhance future data gathering and processing efforts.
Enable users to consider different types of error information.
Improve the reliability of spatial pattern detection by incorporating data quality information and translate uncertainty information into maps.
Communicate data quality information to users.

“I believe that our research will be beneficial to people who use spatially aggregated data, such as cancer rates and survey data, in their analyses,” Chun said. “It is an exciting project to show how analysis results might not be very reliable numerically and visually.”