The correlation dimension of differenced data
arXiv:comp-gas/9405005 · doi:10.1016/0375-9601(94)90134-1
Abstract
The effect differencing has on the estimated correlation dimension of a nongaussian time series is discussed. Two different methods for generating surrogate data sets are compared. The results suggest that any filtering should be applied to both the raw and surrogate time series, otherwise spurious results can occur.
8 pages uuencoded compressed postscript file (using uufiles script)