STOR-i Seminar: Peter Young

Peter Young, Lancaster University

Friday 24 January 2014, 1200-1300
A54, Postgraduate Statistics Centre Lecture Theatre

Hybrid Continuous-Time Box-Jenkins Model Estimation and Forecasting

Abstract: The Hybrid continuous-time Box-Jenkins (BJ) model is a combination of a continuous-time transfer function (differential equation) model and a discrete-time ARMA noise model, where the latter is introduced to avoid the estimation problems associated with the estimation of continuous-time ARMA processes. The advantages of continuous-time modelling from sampled data will be discussed and the optimal Refined Instrumental Variable (RIVC) method of estimating such models will be outlined. The associated estimation algorithm has been implemented as a routine in the CAPTAIN Toolbox for Matlab and it has been applied to diverse practical examples in areas ranging from ecology, through environmental science to engineering. A recent example is the RIVC identification and estimation of the dynamic relationship between total radiative 'forcing' and globally averaged temperature variations, as used in the study of global climate change. The RIVC estimation results will be discussed, including the use of the model for forecasting possible future changes in the globally averaged temperature.

Profile: Peter Young is Emeritus Professor of Environmental Systems at Lancaster University and Adjunct Professor at the Australian National University (ANU). Before his appointment as Head of the Environmental Science Department at Lancaster in 1981, he was a Lecturer in Engineering and Fellow of Clare Hall Cambridge and Professorial Fellow at the ANU. In 1987, he proposed and helped establish the Institute of Environmental and Biological Sciences. He directed the Centre for Research on Environmental Systems for many years and helped to set up the Lancaster Centre for Forecasting with Prof. Robert Fildes in 1992. He has published over 250 papers and book chapters, as well as writing and editing several books, including latterly: Recursive Estimation and Time Series Analysis (2011) and True Digital Control (2013), the latter co-authored with James Taylor and Arun Chotai. A Festschrift for Peter, System Identification, Environmental Modelling and Control System Design, was published in 2012 and he continues to pursue research on data-based mechanistic modelling, forecasting and control.