Abstract: We propose a novel Bayesian methodology to consolidate dynamic sparsity and shrinkage to a unified coherent framework. It combines stochastic variable selection and dynamic shrinkage process in multiple Markov switching processes. The former serves as a dynamic sparsity mechanism by zeroing out model parameters. The latter allows for model parameters to change flexibly or be stable for some time. Our new method achieves a very flexible model dynamics that accommodate zero coefficients, graduation, and sudden structural change in a time-varying and data-driven fashion. Simulation and an application to exchange rate forecasting show the usefulness of this new approach.
The first paper integrates the hierarchical Markov-switching model with model-based clustering to address high-dimensional data challenges, aiming to identify bull and bear markets. This approach enhances both identification and forecasting performance.
Assessing the effect of the ACTN3 gene on athletic performance. Developing a widely applicable general model for studying different genetic effects. The poster can be viewed HERE.
Extending a demographic model initiated by Salop generates some suggestive theoretical results.
• 18th International Joint Conference on Computational and Financial Econometrics and Computational and Methodological Statistics (CFE-CMS2024), King's College London, 2024
• Econometric Society Australasia Meeting (ESAM), Monash University, Melbourne, 2024
• 6th Time Series and Forecasting Symposium, University of Sydney, 2024
• Melbourne Metrics Workshop, The University of Melbourne, 2024
• 2024 FBE Celebrating Graduate Research, The University of Melbourne, 2024
Zheng Fan — fan [dot] z [at] unimelb [dot] edu [dot] au