Resampling chaotic time series

A method for resampling time series generated by a deterministic chaotic data generating process (DGP) is proposed. Given an observed time series, this method potentially allows one to obtain an arbitrary number of time series of arbitrary length which can be considered as a product of the same unknown DGP. The notion of shadowing and brittleness of the pseudo-orbit proves to be particularly useful in characterizing the conditions for a correct resampling. A simple practical application of the method is shown.