Detection of chaotic determinism in stochastic short time series

We have developed an algorithm based on the nonlinear autoregressive (NAR) model which is very accurate in determining whether chaotic determinism is present in a noisy time series and is effective even for a time series with as few as 500 data points. The algorithm is based on fitting a deterministic and stochastic nonlinear autoregressive (NAR) model to the time series, followed by an estimation of the Lyapunov exponents of the resultant fitted model. The major benefits of this algorithm are: 1) it provides accurate parameter estimation with as few as 500 data points, 2) it is accurate down to signal-to-noise ratios (SNRs) of -9 dB (variance of the noise is approximately 2.9 times greater than the variance of the signal), and 3) it allows characterization of the dynamics of the system, and thus prediction of future states of the system. The advantages of the developed algorithm allow this method to be superior to the conventional algorithms for calculating Lyapunov exponents.