A comparative analysis of calibration methods for the SEIR influenza model using synthetic and surveillance data, and extending the SIR model to assess the impact of misinformation
Mumtaz, Nabeela
Mumtaz, Nabeela
Loading...
Repository DOI
Publication Date
2024-11-28
Type
doctoral thesis
Downloads
Citation
Abstract
System Dynamics is a computer-aided approach to strategy and policy design aimed at understanding complex, dynamic systems. Despite its utility, unresolved questions remain regarding model validation, calibration, and structural analysis. Our contributions to the field include:
• First, we explored and compared bootstrap and Hamiltonian Monte Carlo (HMC) methods across various dimensions for parameter estimation in the SEIR model, highlighting their respective strengths and weaknesses. The bootstrap method can be adapted for estimating parameters, although it is computationally intensive. In contrast, HMC offers efficiency, precision, and faster convergence rates.
• Second, a case study on the calibration of the SEIR model using seasonal influenza surveillance data demonstrates that the HMC method provides better model fit and reliability compared to the bootstrap method. This is evidenced by lower MASE values, more precise results, and quicker convergence rates. This study also emphasizes the role of both synthetic and real data in the understanding of statistical methods.
• Third, a case study on the comparative assessment of the SEIR and our SEIR-D compartmental models. The SEIR-D incorporating a time delay to enhance the accuracy of disease spread predictions. This assessment evaluates the impact of seasonal onset delays on model dynamics and improves the estimation of infectious disease parameters crucial for outbreak monitoring and vaccination strategies.
• Fourth, we extended the SIR model structure, introduced two interacting contagion SIR models and applied the Loops That Matter (LTM) method for model structural analysis. This analysis investigated the influence of misinformation on infectious disease transmission dynamics, highlighting its significant impact on disease attack rates and emphasizing the need for targeted interventions to mitigate its effects.
Through these contributions, this research advances System Dynamics by proposing methodologies for model calibration and structural analysis, thereby improving model reliability.
Funder
Publisher
University of Galway
Publisher DOI
Rights
Attribution-NonCommercial-NoDerivatives 4.0 International