Fast detection of nonlinearity and nonstationarity in short and noisy time series

TL;DR

This paper presents a fast, simple, and effective statistical method for detecting nonlinearity and nonstationarity in short, noisy time series, outperforming existing methods in confidence and efficiency.

Contribution

The authors introduce a new detection method that requires only one parameter, is easier to implement, and provides higher confidence in identifying nonlinearity and nonstationarity.

Findings

Method has discrimination power comparable to advanced estimators.

Effective on short, noisy time series.

Rejects null hypothesis with higher confidence than existing methods.

Abstract

We introduce a statistical method to detect nonlinearity and nonstationarity in time series, that works even for short sequences and in presence of noise. The method has a discrimination power similar to that of the most advanced estimators on the market, yet it depends only on one parameter, is easier to implement and faster. Applications to real data sets reject the null hypothesis of an underlying stationary linear stochastic process with a higher confidence interval than the best known nonlinear discriminators up to date.