Identifiable Markov Switching Models with Instantaneous Effects and Exponential Families
Researchers have developed a new framework called FlowMSM to address the challenges of identifying latent regimes and causal structures in non-stationary time series data. This framework is designed to handle complex dynamics, including nonlinear and non-Gaussian behaviors, as well as instantaneous effects between variables. The approach establishes theoretical identifiability for both latent regimes and regime-dependent causal structures, and has demonstrated effectiveness on synthetic benchmarks and a financial economics dataset. AI
IMPACT Provides a new method for analyzing complex time series data, potentially improving applications in finance, climate science, and healthcare.