Testing for general dynamical stationarity with a symbolic data compression technique

We construct a hypothesis test for examining the stationarity of the evolution law for a time series of discrete symbols: whether two data streams appear to originate from the same underlying, but unknown, dynamical system. Based on techniques from the theory of data compression, our method intelligently accounts for the substantial serial correlation and nonlinearity found in realistic dynamical data. We demonstrate the method on a number of realistic experimental datasets.