|
Abstract:
|
In this work, we investigate the problem of multiple change-point detection where the changes occur in the mean of univariate data. The proposed methodology combines the multiscale MOving SUM (MOSUM) procedure and localised pruning based on an information criterion. The MOSUM procedure returns change-point estimates with additional information, such as the local environment in which the change-points are detected, p- values from the asymptotic null distribution and size of jumps at the change-points, which is utilised at the localised pruning stage. The combined approach achieves consistency in estimating the total number and locations of the change-points, is computationally efficient, and shows good empirical performance.
This is joint work with Claudia Kirch and Alexander Meier (Otto-von-Guericke University Magdeburg).
|
