IS.16

IS.16: Exploring the Limits of Multivariate and High Dimensional Inference Methods to Analyze Data i

Motivation:
Ecological data typically consist of more than one measurement from a certain species and questions of interest cover the spatial spread or pattern, survival or viability of the species. Moreover, analyzing multiple species jointly is a necessity due to the diversity of ecosystems. Thus, multivariate data occurs in a natural way and its adequate analysis may even be hampered by high-dimensionality.

In this framework the Arne Bathke (one of the suggested speakers) and the session organizer have a joined D-A-CH-Lead-Agency project on

“Inference methods for multivariate and high-dimensional data”

 which is funded from both, the Austrian and German Research Foundations. Within this project Arne Bathke started a collaboration on quantifying the overlap of niches in which it turned out that nonparametric rank-methods can present a robust and very intuitive solution.  Similarly, the organizer of this session also agreed on a collaborative project on identifying footprints of endangered wild animals which will start in summer. If successful and wished by the organizing committee, the organizer could offer the option to complement this session by an additional talk.

In any case, the main motivational reasons are the dissemination of existing results, further education in ecological and multivariate statistics and discussion with experts from the field.

Organizer:
Markus Pauly, University of Ulm, Institute of Statistics

Speakers:
Marti J. Anderson, New Zealand Institute for Advanced Study (NZIAS), Massey University
Modelling nonlinear responses of species to environmental gradients along with species' associations in ecological communities - some parametric and nonparametric approaches

Arne C. Bathke, Department of Mathematics, University of Salzburg and Department of Statistics, University of Kentucky
Quantifying ecological niches, with confidence

Mikoko Minami, Department of Mathematics, Keio University
Analysis of count data with many zero-valued observations: over-estimation of trend by negative binomial regression

Discussant:
Frank Konietschke, Charite University Medical Center Berlin