Inference on Classical Change Point Problem

Gaur, Naveen Prakash (2017) Inference on Classical Change Point Problem. Masters thesis, Indian Institute of Science Education and Research Kolkata.

PDF (MS dissertation of Naveen Prakash Gaur (12MS093)) 12MS093_MS_Thesis.pdf - Submitted Version Restricted to Repository staff only Download (357kB)

Abstract

Irrespective of being an age old problem, classical change point problem draws the attention of a statistician for its wide applications. The maximum likelihood method is mostly used for detection of a change point in this setup. But it is interesting to note that maximum likelihood method and CUSUM method both give a false alarm of the existence of change point even though there is no change point at all in the given sequence of the data. We have studied mean change problem of Gaussian sequence for a known and an unknown variances. Apart from the maximum likelihood estimator, we use other plug-in estimators of the variance in the estimation and the testing problems related to the change point. We find that the newly proposed testing procedure is at par or even better than the maximum likelihood based testing procedure. Similar results are also observed for multiple change-point problems.

Actions (login required)

View Item