bayesian-sgdlm is a Python script for fully Bayesian SGDLMs, treating each node as a VAR( 𝑝) DLM. It leverages decouple–recouple filtering with Variational Bayes and importance sampling to estimate sparse, time-varying cross-lag dependencies (including pandemic dummies) without ever inverting the full multivariate system.
reproducible-research variational-bayes importance-sampling dynamic-linear-model financial-econometrics state-space-model multivariate-time-series bayesian-time-series sgdlm sequential-forecasting
-
Updated
Aug 10, 2026 - Python