Geographically weighted temporally correlated logistic regression model

Area of Science:

• Epidemiology
• Biostatistics
• Geospatial analysis

Background:

• Understanding dynamic correlations in biological and disease systems is crucial.
• Existing models may not fully capture spatio-temporal variations.

Purpose of the Study:

• To propose a novel model, geographically weighted temporally correlated logistic regression (GWTCLR), for identifying dynamic correlations in binomial outcome data.
• To incorporate spatial and temporal information for joint inference of predictor relationships.

Main Methods:

• The GWTCLR model utilizes local likelihood for spatial relationship estimation.
• Temporal variations are estimated using smoothing methods.
• Asymptotic properties of the estimator are analyzed.

Main Results:

• Simulation studies demonstrate the robustness of the GWTCLR model.
• Application to seasonal influenza data revealed previously unobservable spatial and temporal varying patterns.
• Results were largely consistent with previous epidemiological studies.

Conclusions:

• The GWTCLR model effectively identifies dynamic correlations in spatio-temporal data.
• It offers enhanced insights into complex epidemiological patterns.
• The model provides a valuable tool for analyzing disease determinants.