Exponential Smoothing and State-Space Modeling

Exponential smoothing methods for forecasting are sometimes regarded as a relic of the 1960s or, at best as a special case of simple ARIMA models. Neither characterization (or caricature) is correct.

In this presentation we will examine the linkage between exponential smoothing and state-space models and show how this connection opens up ways to deal, at least approximately, with non-linear processes. Extensions to heteroscedastic processes are readily incorporated into the framework and several examples will be presented, including analyses of gasoline prices and of the records of an aging runner.

Dr. Keith Ord
McDonough School of Business,
Georgetown University

Bio

Professor Ord's interests lie in business forecasting and statistical modeling and analysis of business processes. His most recent book with R.J. Hyndman, A.B. Koehler and R.D. Snyder, is titled Forecasting with Exponential Smoothing: The State Space Approach, Springer 2008. He is also a co-author of Kendall's Advanced Theory of Statistics, a two-volume reference work now in its sixth edition, and several books on spatial statistics and time series. Professor Ord has been a professor with McDonough School of Business, Georgetown University since 1999. His prior positions include professorship and chairman at Penn State University and at the University of Warwick, England. He is a fellow of the American Statistical Association and of the International Institute of Forecasters. He is also an elected member of the International Statistical Institute.