A0878
Title: Bayesian bi-directional self-exciting threshold autoregressive model and the application in loss reserving
Authors: Yuning Zhang - The University of Sydney Business School (Australia) [presenting]
Boris Choy - University of Sydney (Australia)
Wilson Ye Chen - University of Technology Sydney (Australia)
Tak Kuen Siu - Macquarie University (Australia)
Abstract: A Bayesian statistical approach is proposed for estimating a self-exciting threshold autoregressive model (SETAR) in bidirectional time series (bi-SETAR). While the frequentist SETAR, adapted into a bidirectional framework (SETAR-NN), has recently been utilized for claim reserving in run-off triangles, the proposed Bayesian approach introduces a more flexible and practical methodology for analyzing and estimating the SETAR-NN model. This approach focuses on providing probabilistic estimates for the structural parameters, emphasizing the threshold parameters. The Markov Chain Monte Carlo (MCMC) method is employed to simulate the posterior distributions of unknown parameters and predictive distributions. Applications in loss reserving and computing risk metrics are also demonstrated, and they are compared against the results from benchmark models.
