A1107
Title: Non-homogeneous Markov-switching generalized additive models for location, scale, and shape
Authors: Timo Adam - Bielefeld University (Germany) [presenting]
Katharina Ammann - Bielefeld University (Germany)
Abstract: The aim is to propose an extension of Markov-switching generalized additive models for location, scale, and shape (GAMLSS) that allows covariates to influence both the parameters of the state-dependent distributions and the transition probabilities. Traditional Markov-switching GAMLSS combines distributional regression with latent-state time series modelling, but typically assumes constant transition probabilities, which prevents regime shifts from responding to covariate-driven changes. The approach overcomes this limitation by allowing the transition probabilities to vary with covariates, thereby capturing covariate-dependent regime dynamics. The proposed methodology is evaluated through simulation studies, and its practical usefulness is illustrated in a case study on Spanish energy prices.
