Colloquium Speaker:
Dr. Severien Nkurunziza
University of Windsor
Title: Robust inference and change-point methods in generalized mean-reverting processes
Abstract: In this talk, we present a comprehensive overview of recent results on change-point analysis and improved inference methods for generalized mean-reverting processes. The proposed processes are particularly well-suited for data exhibiting a non-constant mean-reverting level, along with seasonal trends and unconventional shocks. We also study the inference problem concerning the drift parameter of the proposed processes, specifically in contexts where the target parameter is likely to satisfy certain restrictions. We develop three types of estimators for the drift parameter: the unrestricted estimator, the restricted estimator, and a class of shrinkage estimators. Furthermore, we analyze the relative efficiency of the proposed estimators under a class of local alternative restrictions. We also construct an asymptotic test for hypotheses concerning the drift parameter. In addition to assessing potential restrictions that may constrain components of the drift parameter, the proposed test is particularly effective in detecting the absence of a change-point. Moreover, we establish the test’s asymptotic power and prove its consistency.
Day & Time: Friday, October 31, 2025, at 3:00pm
Location: Lambton Tower, Room 9-118