Local indicators of geocoding accuracy (LIGA): theory and application

Area of Science:

• Spatial statistics
• Geographic Information Systems (GIS)
• Geospatial analysis

Background:

• Positional error in geographic locations is a known issue in spatial analysis.
• Limited research exists on how this error affects statistical outcomes.
• This study addresses the need to quantify the impact of positional error on spatial weights.

Purpose of the Study:

• To develop and evaluate methods for quantifying the perturbability of spatial weights in response to positional error.
• To assess the sensitivity of spatial statistics to varying levels and types of positional error.
• To identify geographic locations most affected by positional error for targeted accuracy improvements.

Main Methods:

• Exploration of techniques to quantify the perturbability of spatial weights.
• Development of a family of curves to describe the relationship between perturbability and positional error.
• Evaluation of spatial weight specifications globally and locally.
• Simulation studies and application to a bladder cancer case-control study.

Main Results:

• A family of curves effectively models the relationship between positional error and spatial weight perturbability.
• The mean positional error, not the distribution shape, significantly influences perturbability.
• A novel visualization tool, the LIGA scatterplot, identifies sensitive locations.

Conclusions:

• Methodology enables routine sensitivity analysis of spatial statistics to positional error from various sources (e.g., GPS, geocoding).
• High-perturbability, high-leverage locations benefit most from increased positional accuracy.
• Understanding positional accuracy is crucial for validating spatial analysis results.