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

Download:

Current browse context:

stat.ME

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

Statistics > Methodology

Title: Mixtures of Contaminated Matrix Variate Normal Distributions

Abstract: Analysis of three-way data is becoming ever more prevalent in the literature, especially in the area of clustering and classification. Real data, including real three-way data, are often contaminated by potential outlying observations. Their detection, as well as the development of robust models insensitive to their presence, is particularly important for this type of data because of the practical issues concerning their effective visualization. Herein, the contaminated matrix variate normal distribution is discussed and then utilized in the mixture model paradigm for clustering. One key advantage of the proposed model is the ability to automatically detect potential outlying matrices by computing their \textit{a posteriori} probability to be a "good" or "bad" point. Such detection is currently unavailable using existing matrix variate methods. An expectation conditional maximization algorithm is used for parameter estimation, and both simulated and real data are used for illustration.
Subjects: Methodology (stat.ME)
Cite as: arXiv:2005.03861 [stat.ME]
  (or arXiv:2005.03861v1 [stat.ME] for this version)

Submission history

From: Michael Gallaugher [view email]
[v1] Fri, 8 May 2020 05:32:18 GMT (16kb)

Link back to: arXiv, form interface, contact.