Talk 3

One Day Meeting on Advanced Statistics Topics

CIDMA- University of Aveiro

Speaker: Peter Filzmoser - TU Wien, Austria

Title of the talk: Robust Statistical Methods for Matrix- and Tensor-Valued Observations and Random Surfaces

Abstract: Modern measurement techniques and devices produce observations in many forms: vectors, matrices (such as images), functions (such as spectra), and more complex structures. In this talk, we present robust statistical approaches for matrix- and tensor-valued data, as well as for functional data observed as surfaces. For random surfaces with a separable covariance structure, we show that the basis function representation follows a matrix-variate distribution, which links the two settings. We introduce robust estimators for the model parameters and provide tools to explain why individual observations are flagged as outlying. Applications to real data demonstrate the practical value of these methods.

Peter Filzmoser is a Professor of Statistics at TU Wien, Austria, where he heads the research unit Computational Statistics. His research interests include robust statistics, multivariate analysis, compositional data analysis, and geostatistics. He has authored more than 200 research articles and several R packages, and is a co-author of books on compositional data analysis (Springer, 2018), multivariate methods in chemometrics (CRC Press, 2009), and analyzing environmental data (Wiley, 2008).

ORCID: 0000-0002-8014-4682

Page updated

Google Sites

Report abuse