tseriesChaos is an R package for the analysis of nonlinear time series mainly motivated by chaos theory.
You can install from a CRAN mirror near you by typing, at the R prompt:
> install.packages('tseriesChaos', dep=TRUE)
Its developement was initially inspired by the TISEAN project.
Currently, the 0.1-7 version is out, and mainly includes explorative methods. Among others:
- Kantz method for Maximal Lyapunov Exponent (MLE) estimation
- efficient correlation integral computation for estimating fractal dimension
- false nearest neighbours
- space-time separation plots
- recurrence plots
- grid-based average mutual information computation
Forthcoming
A new, completely revised and extended version of the package supported by the PRIN project coordinated by Estela Bee Dagum, Statistics Department, University of Bologna.
The new version will include:
- Statistical tests for chaos based on Lyapunov exponents estimation with different methods available:
- Kantz method, resampled version (Giannerini and Rosa, Physica D (2001));
- Neural Network, asymptotic version (Shintani and Linton, J. Econometrics (2005));
- Local Lyapunov exponents estimation
- Extension to the stochastic case of Lyapunov Exponents: measures of sensitivity to initial conditions of the conditional mean (Yao and Tong JRSS B (1994)) and of the conditional distribution (Fan, Yao and Tong, Biometrika (1996)), see also Giannerini and Rosa SNDE (2004).
- Entropy based tests for nonlinearity (Granger, Maasoumi & Racine JTSA 2004, Giannerini, Maasoumi & Bee Dagum forthcoming).
