Comparison of Bayesian Spatiotemporal Models for Small-Area Life Expectancy: A Simulation Study

Area of Science:

• Biostatistics
• Demography
• Spatial Analysis

Background:

• Accurate small-area life expectancy estimation is crucial for public health.
• Bayesian spatiotemporal models are increasingly used for complex demographic analyses.
• Assessing model performance, including precision, uncertainty, and distributional assumptions, is vital.

Purpose of the Study:

• To evaluate the precision, uncertainty, and normality of life expectancy estimates from Bayesian spatiotemporal models.
• To compare these models against traditional and Bayesian spatial methods.
• To determine an optimal cutoff value for statistical significance in life expectancy comparisons.

Main Methods:

• Simulated 6 population size scenarios (500-25,000) for 247 South Korean districts (2013-2017).
• Generated 1,000 hypothetical datasets per scenario.
• Calculated district-year life expectancies using Bayesian spatiotemporal, Bayesian spatial, and traditional methods.
• Assessed precision, 95% uncertainty interval coverage, and normality of estimates.

Main Results:

• Bayesian spatiotemporal models yielded precise life expectancy estimates.
• Uncertainty intervals did not consistently contain the true value with 95% probability.
• Normality assumption was violated in small population scenarios.
• A cutoff value of 0.8 is proposed to minimize false positives/negatives.

Conclusions:

• Bayesian spatiotemporal models provide precise estimates but require careful interpretation regarding uncertainty and normality, especially for small populations.
• The proposed 0.8 cutoff value aids in statistically significant life expectancy comparisons.
• Further research should validate these findings across diverse populations and settings.