A multivariate generalized logistic approach with spatially varying nonlinear components for modeling epidemic data.

Area of Science:

• Epidemiology
• Biostatistics
• Spatial Statistics

Background:

• Traditional epidemiological time series analysis often focuses on linear predictors for mean outcomes.
• Existing hierarchical models may not fully capture non-linear patterns in epidemic counts across regions.
• There's a need for methods that account for spatial similarities in non-linear epidemiological trends.

Purpose of the Study:

• To propose and evaluate a novel spatial specification for non-linear components in hierarchical models for epidemiological count data.
• To improve parameter estimation and prediction of future epidemic outcomes in neighboring regions.
• To apply the proposed methodology to real-world data, including COVID-19 case counts.

Main Methods:

• Development of a hierarchical model incorporating spatial specifications for non-linear components.
• Parametric modeling of epidemic waves using a data-driven approach, considering multiple waves.
• Testing the model through simulation studies and application to real epidemiological data.

Main Results:

• The proposed spatial model demonstrates strong fitting and prediction capabilities.
• Simulation studies validate the model's performance.
• Application to COVID-19 case data shows favorable comparison against alternative epidemiological models.

Conclusions:

• The spatial specification for non-linear components offers a valuable enhancement to epidemiological time series analysis.
• The methodology provides a robust framework for joint analysis and prediction of count data in neighboring regions.
• The approach is effective for modeling complex epidemic dynamics, such as multiple waves.