Program package used for estimating common trends, seasonal components and cycles in short, non-stationary multivariate time series. Based on dynamic factor analysis.
Based on the model-free method of time series analysis Caterpillar-SSA (Singular Spectrum Analysis). The result of the Caterpillar-SSA processing is identification, analysis and forecast of additive components of time series (trends, periodicities, noise). The program can be applied to multivariate analysis/forecasting and change-point detection.
Using novel tools such as singular spectrum analysis and multitaper method, time series can be decomposed into noise and predictable components - trends and oscillatory modes, which can be reconstructed and forecast.