This paper introduces a novel framework for estimating Time-Varying Exposure-Response Functions (TVERF) within non-stationary diffusion dynamics, integrating distributed lag non-linear models with time-varying coefficient estimation through a stochastic differential equation framework. The approach uses Bayesian estimation with Gaussian process priors to account for temporal heterogeneity in exposure effects and non-stationary diffusion coefficients.
Key findings
Develops a unified mathematical framework combining distributed lag non-linear models with non-stationary diffusion processes.
Proposes a Bayesian estimation procedure using Gaussian process priors for posterior inference.
Establishes theoretical properties including consistency and asymptotic normality of the proposed estimators.
Limitations & open questions
Further research is needed to extend the framework to other types of environmental exposures and health outcomes.