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: A convex approach to optimum design of experiments with correlated observations

Abstract: Optimal design of experiments for correlated processes is an increasingly relevant and active research topic. Present methods have restricted possibilities to judge their quality. To fill this gap, we complement the virtual noise approach by a convex formulation leading to an equivalence theorem comparable to the uncorrelated case and to an algorithm giving an upper performance bound against which alternative design methods can be judged. Moreover, a method for generating exact designs follows naturally. We exclusively consider estimation problems on a finite design space with a fixed number of elements. A comparison on some classical examples from the literature as well as a real application is provided.
Subjects: Statistics Theory (math.ST); Methodology (stat.ME)
MSC classes: 62K05
Cite as: arXiv:2103.02989 [math.ST]
  (or arXiv:2103.02989v3 [math.ST] for this version)

Submission history

From: Werner Müller [view email]
[v1] Thu, 4 Mar 2021 12:29:46 GMT (238kb,D)
[v2] Wed, 20 Oct 2021 16:36:42 GMT (541kb)
[v3] Fri, 22 Oct 2021 10:50:48 GMT (541kb)

Link back to: arXiv, form interface, contact.