Robust Statistics and Extremes

 In Workshop reports

Australian National University, 8–11 September 2014

The Robust Statistics and Extremes Conference brought the related but not usually connected themes of robustness and extremes together with functional data analysis in a stimulating and enjoyable event. Australia has a significant presence on the world stage in the important and highly topical areas of robust statistics and extreme value theory, and their applications. There have been enormous advances in recent years in the application of these methods in particular to the analysis of extremely large data sets and high dimensional data such as is found in genetics, finance, physics, astronomy, and many other areas.

The conference featured a series of presentations, seminars and practical discussions exploring the cutting-edge of robust statistics, extreme value theory and its applications, and high dimensional and function data analysis.

The robust statistics sessions showcased the very latest developments in the field. A panel of experts presented research on high-dimensional data analysis, robust filtering, outlier detection for functional data and robust component analysis, generalised likelihood ratio tests for heterogeneous sparse normal mixtures, and one-sided Winsorization sample surveys. Presenters included Peter Hall (The University of Melbourne), Elvezio Ronchetti (University of Geneva), Matias Salibian-Barrera (University of British Columbia), Jingling Wu (University of Calgary), Janice Scealy (Australian National University), Thomas Porter (The University of Sydney) and Robert Clark (University of Wollongong).

With a focus on practical discussion, the extreme value sessions touched on the ever-topical and critical issue of climate change, as well as other modelling applications and theoretical presentations. Presenters included Ana Ferreira (ISAUL and CEAUL, Portugal), Debbie Dupuis (HEC Montreal), Yuguang Fan (Australian National University), Daniel Gibbons (The University of Queensland), Yang Liu (CSIRO), Wangyue Xie (The University of Queensland) and Peter Cayton (Australian National University).

High Dimensional and Functional Data Analysis is currently a very hot topic of research related to interests in “Big Data” and “Machine Learning”. Seminars included those by Aurore Delaigle (The University of Melbourne): Classification Using Censored Functional Data, HanLin Shang (Australian National University): A Bayesian method for determining the optimal semi-metric and bandwidth in scalar-on-function quantile regression with unknown error density and dependent functional data, Robert Staudte (LaTrobe University): Inference for Ratios of Finite Linear Combinations of Quantiles, and Garth Tarr (Australian National University): A pairwise approach to estimating precision and covariance matrices under cell-wise contamination.

On the final morning a Swiss Tea was a highlight. It was held to mark the 50th anniversary of the publication of P.J.Huber’s 1964 paper, Robust estimation of a location parameter. This paper played a major role in the development of robustness theory, as it introduced M-estimators, established their asymptotic normality (including in non-smooth cases), proposed and implemented a minimax treatment of the robust estimation, derived the least favourable distribution for the core Gaussian location problem and the associated Huber M-estimator, showed the optimality of the median for asymmetric contamination, proposed three methods of simultaneously estimating scale (including the famous Huber proposal II method), and more.

The conference was very well received, bringing together prominent Australian and international experts to provide researchers and students with the opportunity to hear up-todate, state-of-the-art expositions. There was a substantial audience of local and international researchers and postgraduate students in attendance. The conference also attracted significant interest from government and industry

MathSciNet Classification
62G32, 60G70, 60K10

Program structure

Web Links
http://maths.anu.edu.au/events/robust-statistics-andextremes

Other Sponsors
Mathematical Sciences Institute “Special Year”, the Australian National University Research School of Finance and Actuarial Studies and Statistics, the Australian National University

Contact
Professor Ross Maller, Australian National University
ross.maller@anu.edu.au

Organisers

Professor Ross Maller, Australian National University
Dr Boris Buchmann, Australian National University
Yuguang Fan, Australian National University
Dr Bronwyn Loong, Australian National University
Professor Steven Roberts, Australian National University
Dr Dale Roberts, Australian National University
Professor Alan Welsh, Australian National University
Tiandong Wang, Australian National University
Professor Ana Ferreira, University of Lisbon

Special presenters

Professor Aurore Delaigle, The University of Melbourne
Research interests: functional data classification
Professor Debbie Dupuis, University of Montreal
Research interests: robustness and extreme values, with applications in areas ranging from temperature modelling to futures derivatives and actuarial studies
Professor Manuel Febrero-Bande, University De Santiago de Compostela
Research interests: aspects of generalised additive models, with special reference to functional models and data, empirical processes, influence and outliers, etc. Applications are to times series, ozone data, financial data, etc
Professor Ana Ferreira, University of Lisbon
Research interests: extreme value methods theory and its application
Professor Peter Hall, The University of Melbourne
Research interests: nonparametrics and semiparametrics, and applications in economics and in the physical, engineering and biological sciences
Dr Luke Prendergast, La Trobe University
Research interests: robust statistics, dimension reduction and visualisation of high dimensional data, meta analysis (in particular meta regression) and statistical analysis for weight loss studies
Professor Elvezio Ronchetti, University of Geneva
Research interests: theory of robust statistics and the application of robust procedures to statistics, econometrics, and finance, asymptotic expansions, saddlepoint techniques, exploratory data analysis and statistical computing
Professor Matias Salibian-Barrera, University of British Columbia
Research interests: robustness and bootstrapping

Participant Summary

Attendees

NUMBER OF ATTENDEES

XX

  • Academic

    64%

  • Postdocs

    0%

  • Post Graduates

    28%

  • Students

    8%

  • Industry

    0%

Gender

Gender Breakdown

State Breakdown
Victoria
50%
Australian Capital Territory
50%
New South Wales
50%
Queensland
0%
South Australia
0%
Western Australia
0%
Tasmania
0%
Northern Territory
0%
International
0%
Citizenship Status
Domestic - 60%
50%
International - 40%
50%
Recent Posts