An Approach to Time Series Modeling and Forecasting Illustrated by Hourly Electricity Demands.
STATE UNIV OF NEW YORK AT BUFFALO AMHERST STATISTICAL SCIENCE DIV
Pagination or Media Count:
Part I of this paper defines the time series modeling problem in terms of whitening filters. For stationary time series, approximate autoregressive representations play a central role. For non-stationary time series, the modeling problem is to interpret a whitening filter as a series of filters in tandem which provide de-trending and seasonal adjustment procedures. The CAT criterion to determine the optimum order to autoregressive approximation to a sample of size T is described. The concepts of non-predictable and predictable time series, and naive predictors, are introduced and used to help yield models of observed time series suitable for interpretation as well as forecasting. Part II discusses an empirical time series analysis of a long time series of hourly electricity demand.
- Statistics and Probability
- Electricity and Magnetism