We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

math.ST

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Statistics Theory

Title: Cluster based inference for extremes of time series

Abstract: We introduce a new type of estimator for the spectral tail process of a regularly varying time series. The approach is based on a characterizing invariance property of the spectral tail process, which is incorporated into the new estimator via a projection technique. We show uniform asymptotic normality of this estimator, both in the case of known and of unknown index of regular variation. In a simulation study the new procedure shows a more stable performance than previously proposed estimators.
Subjects: Statistics Theory (math.ST); Methodology (stat.ME)
MSC classes: 62G32, 62M10, 62G05, 60F17
Cite as: arXiv:2103.08512 [math.ST]
  (or arXiv:2103.08512v1 [math.ST] for this version)

Submission history

From: Holger Drees [view email]
[v1] Mon, 15 Mar 2021 16:36:25 GMT (577kb,D)

Link back to: arXiv, form interface, contact.