时间：2014年3月14日（周五） 下午3:40

地点：浦江办公楼408会议室

内容：Bayesian time series and forecasting

主讲人：Kostas教授 英国谢菲尔德大学

Abstract:

In this talk I wil go through some of the important areas of time series analysis and Bayesian forecasting. I will start by giving a general background and some historical comments of the discipline and then I will describe several motivating examples, including time series decompositions, multinomial modelling, multivariate volatility estimation, and graphical models. The talk will give an overview of the methods highlighting the possibilities and advantages of the Bayesian paradigm, rather than going into technical details of any method concerned. Relevant methodology such as the Kalman filter and its variants and particle filters will be discussed as necessary in each example. Some methodological challenges will be picked on the go, but the purpose of the talk will have an applied focus. All examples are illustrated with real data from economics/finance and genetics. The talk will conclude on some challenges for the future.