Need to model multivariate time series data? ARfit, a collection of MATLAB modules, offers a robust solution for modeling and analyzing such data using autoregressive (AR) models. This software package includes tools for parameter estimation, model analysis, and AR process simulation. ARfit utilizes a stepwise least squares algorithm to efficiently estimate AR model parameters, especially for high-dimensional data. The modules construct confidence intervals for estimated parameters and compute statistics to assess model adequacy. Dynamical characteristics are examined via decomposition into eigenmodes, oscillation periods, damping times, and excitations. Approximate confidence intervals are provided for eigenmodes and their oscillation periods and damping times. ARfit empowers researchers and practitioners to effectively model, analyze, and simulate multivariate time series data, enabling deeper insights into complex dynamic systems. Its computationally efficient algorithms and comprehensive analysis tools make it a valuable resource for various applications, including signal processing, financial analysis, and climate modeling.
Published in ACM Transactions on Mathematical Software, this article describes a valuable software package for mathematical modeling and analysis. By providing a detailed overview of the ARfit algorithm and its functionalities, this paper contributes to the journal's focus on disseminating high-quality mathematical software and algorithms for various scientific and engineering applications.