Detection of abrupt changes in statistical models
This dissertation investigates different types of disorder problems by using sequential procedures for on-line implementation. The problem is considered within the framework of detecting abrupt changes in an observed random process when the disorder can occur at unknown times. The focus of this work...
| Main Author: | |
|---|---|
| Other Authors: | |
| Language: | en_US |
| Published: |
Monterey, California. Naval Postgraduate School
2014
|
| Online Access: | http://hdl.handle.net/10945/43764 |
| Summary: | This dissertation investigates different types of disorder problems by using sequential procedures for on-line implementation. The problem is considered within the framework of detecting abrupt changes in an observed random process when the disorder can occur at unknown times. The focus of this work is on quickest detection methods for cumsum procedures implemented for different parametric and nonparametric nonlinearities and their performance evaluation. Both the non-Bayesian (Maximum-Likelihood) and the Bayesian frameworks are presented but the focus is mainly on non-Bayesian methods for which detailed analysis is provided. The use of Brownian motion approximations is also included and provides an additional viewpoint of analyzing the performance for both the non-Bayesian and Bayesian methods. |
|---|