Predicting seasonal influenza transmission using functional regression models with temporal dependence

Area of Science:

• Epidemiology
• Biostatistics
• Environmental Health

Background:

• Accurate influenza incidence prediction is crucial for public health resource allocation.
• Traditional methods face challenges with delayed or unavailable recent case data.
• Meteorological data offers a potential alternative for timely influenza forecasting.

Purpose of the Study:

• To develop and evaluate a novel approach for predicting influenza incidence using meteorological information.
• To extend Generalized Least Squares (GLS) methods to functional regression models with dependent errors for influenza forecasting.
• To assess the performance of the proposed model against classical statistical and time series approaches.

Main Methods:

• Application of multivariate functional regression models with dependent errors, extending Generalized Least Squares (GLS).
• Utilizing distance correlation to select relevant meteorological and functional variables for prediction.
• Development of an iterative GLS (iGLS) estimator to handle complex data dependencies.
• Conducting a simulation study to compare GLS estimators with classical models.

Main Results:

• GLS estimators provide superior parameter estimations compared to classical regression models.
• The proposed GLS approach demonstrates excellent predictive performance for influenza incidence.
• The model is competitive with established time series methods for influenza forecasting.
• The iGLS variant effectively models intricate dependence structures.

Conclusions:

• Meteorological information, when integrated with advanced regression techniques, offers a robust method for influenza incidence prediction.
• The developed functional regression models with dependent errors are highly effective, particularly when recent epidemiological data is compromised.
• This approach provides valuable tools for health managers to proactively allocate resources and manage influenza epidemics efficiently.