Time Varying Autoregressive (TVAR) Model Toolbox

This toolbox provides robust implementations for handling Time Varying Autoregressive (TVAR) models. TVAR models are specifically designed to capture and analyze time-dependent autoregressive relationships in dynamic systems. The toolbox incorporates multiple algorithms for model selection criteria like AIC/BIC, parameter estimation through recursive least squares or Kalman filtering techniques, and comprehensive model diagnostic tools including residual analysis and stability checks. Users can efficiently implement sliding window approaches or state-space formulations for time-varying coefficient tracking. The package includes visualization functions for plotting time-evolving coefficients, spectral characteristics, and prediction intervals using MATLAB's graphics capabilities. With built-in functions for handling non-stationary time series data and real-time parameter adaptation, this toolbox enables researchers to perform accurate forecasting and gain deeper insights into dynamic system behaviors. Advanced plotting modules allow users to visualize model convergence, parameter trajectories, and predictive performance metrics through interactive charts and comparative analysis tools. Overall, this serves as an essential resource for implementing and interpreting TVAR models in practical applications.