Advances and Applications in Statistics
Volume 27, Issue 2, Pages 131 - 165
(April 2012)
|
|
DETECTION OF CHANGE POINTS: A SURVEY OF METHODOLOGIES
Arunendu Chatterjee
|
Abstract: An attempt has been made, in this paper, to develop Bayesian change point detection techniques based on different wavelet transformations. For this purpose, we have taken an overview of both frequentist and Bayesian methods. We observe that both the methods have some limitations from the view point of tackling real data which are sometimes characterized by the occurrence of missing values and noise. To overcome this problem, we suggest that wavelet procedures can be used to ignore missing values and speed up the detection process by denoising the data. We also observe that most of the procedures in the change point literature have been examined only on simulated data, and do not address the complications that arise if one intends to use the procedures, in practice. We have provided some empirical illustrations to address the problem on the real dataset. |
Keywords and phrases: change points, discrete wavelet transform, lifting, Bayesian method, real data problems. |
|
Number of Downloads: 345 | Number of Views: 1074 |
|