Extended Martins–Rodrigues Method in Detecting Multiple Persistence Change

The occurrence of multiple persistence changes in long-memory time series data affects both model parameter estimation and forecasting performance. This study presents an extension of the Martins–Rodrigues method for detecting multiple persistence changes in time series processes. The extended method enhances the original framework by enabling the effective and meaningful identification of multiple persistence changes in the time series data. Its performance is evaluated through simulation by comparison with the method proposed by Leybourne et al. The results indicate that the extended method demonstrates strong accuracy and reliability in detecting multiple persistence changes for the increasing values of persistence change parameters, as evidenced by a higher percentage of break-point detection. We then apply the extended method to the Philippines inflation rate data. We demonstrate that the forecasting performance improves after accounting for the detected break point of persistence changes.