Related Experiment Videos
Scale and shape issues in focused cluster power for count data
Robin C Puett1, Andrew B Lawson, Allan B Clark
1Department of Epidemiology and Biostatistics, Arnold School of Public Health, University of South Carolina, Columbia, SC, USA. rpuett@hsph.harvard.edu.
International Journal of Health Geographics
|April 2, 2005
Summary
Statistical power for disease cluster detection varies significantly with cluster shape and scale. Selecting statistical models that account for spatial patterns is crucial for accurate environmental exposure investigations.
Area of Science:
- Environmental epidemiology
- Biostatistics
- Spatial analysis
Background:
- Growing interest in statistical methods for disease cluster detection.
- Need for power evaluations in environmentally-related disease cluster investigations.
- Gaps in existing research regarding focused cluster detection, scale, and shape.
Purpose of the Study:
- To simulate disease clusters of various shapes and scales around a pollution source.
- To evaluate the statistical power of focused cluster tests and generalized linear models for count data.
- To assess the impact of cluster characteristics on detection power.
Main Methods:
- Simulation of disease clusters with diverse spatial patterns (shapes and scales).
- Evaluation of statistical power for various focused cluster detection tests.
- Assessment of generalized linear models for detecting spatial disease patterns.
Main Results:
- Statistical power generally increased when models incorporated shape-specific parameters (e.g., directional functions).
- Detection power was limited for clusters with a peaked then declining risk pattern.
- Significant variations in power were observed based on cluster scale, shape, and the statistical method used.
Conclusions:
- The scale and shape of disease clusters substantially influence the power of statistical detection methods.
- Selecting statistical tests and models with functions tailored to the expected spatial pattern is essential.
- Accurate detection of environmentally-related disease clusters requires careful consideration of spatial characteristics.