AISM 53, 552-566
© 2001 ISM
(Received June 14, 1999; revised November 16, 1999)
Abstract. We suggest a modification of the CUSUM procedure to detect changes in angular data. We obtain limit theorems for the test statistics under the no change null hypothesis. We discuss the estimation of the times of changes and show that the binary segmentation provides the times of all changes. Our method is applied to a data set on the activity of a pulsar.
Key words and phrases: Angular data, change-point, pulsar, Brownian bridge, von Mises distribution.