【Metrics Seminar】Bregman Model Averaging for Generalizing Asymptotically Optimal Combined Forecasts
研討會日期 : 2023-11-21
時間 : 14:30
主講人 : Professor Chu-An Liu
地點 : Conference Room B110
演講者簡介 : Professor Chu-An Liu received his PhD from the University of Wisconsin-Madison in 2012. He is currently an Associate Research Fellow at the Institute of Economics, Academia Sinica. His research interests are Econometric Theory and Applied Econometrics.
演講摘要 : We propose a unified model averaging approach for establishing a wide class of asymptotically optimal combined forecasts. In this approach, the asymptotic optimality is established by minimizing an asymptotic risk based on the expected Bregman divergence of a combined forecast from a generalized forecasting target under the local (to-zero) asymptotics. This approach is flexibly applicable to generate new model averaging methods in various forecasting contexts. Examples include, but are not limited to, univariate or multivariate conditional-mean forecasts, volatility forecasts and probabilistic forecasts. This is an important feature of our approach. As an illustrative example, we utilize this unified approach to generate new model averaging methods for combining vector autoregression forecasts, and compare our methods with related existing methods by a Monte Carlo simulation and an empirical application. The comparison shows that our methods perform reasonably well in finite samples and in real data.