Author(s): H. F. P. Van Den Boogaard; R. E. Uittenbogaard; A. E. Mynett
Linked Author(s):
Keywords: Synthetic Time Series; Statistical Constraints; Marginal Probability Distributions; Correlation and Covariance Functions; Cost Functions; Interpolation/Extrapolation of Time Series
Abstract: An algorithm is described for the construction of synthetic time series. A synthetic time series is a function of time X (⋅) that satisfies one or more a priori prescribed statistical properties. Such statistical constraints may for example be a marginal probability density distribution, and/or an auto-covariance function. Moreover it is taken into account that the synthetic series agrees with observations when and where these are available. Synthetic time series are valuable for the interpolation of missing values or gaps in observed series, or the generation of (ensembles of) fully artificial series with desired statistical properties. Such ensembles provide important opportunities for uncertainty and/or risk analysis. The potential of the proposed method for the construction of synthetic time series is illustrated by applications.
Year: 2003