Thursday 11 Dec 2014: Statisical Science seminar: Analysis and Forecasting of Locally Stationary Time Series
Prof Guy Nason - University of Bristol
Faced with a new time series a statistician has many questions to ask. What kind of models are appropriate? Is the sampling rate appropriate? Is the series stationary? Are nonlinear models appropriate? How can I produce good forecasts? This talk advertises a collection of tools that can provide answers or partial answers to some of these questions in situations where it is suspected that the underlying time series is not stationary. We shall summarize some of the theory underlying these new methods and also demonstrate their performance in simulated situations. Finally, we will show these tools in action on some time series recorded on the economy and from the field of wind energy.