I would like to remind members of the list of a meeting
about Visualisation of Complex Data Sets
organised by the General Applications Section of the
Royal Statisticial Society that will take place on
Wednesday 17th May
at the RSS Headquarters, Errol Street, London
2pm to 5pm (Tea at 3:30pm approximately)
The abstracts of the talks are below.
Thank you very much.
Julian Stander
GAS Secretary
---------------------------------------------------------------------------
Visualisation of Complex Data Sets
ANTONY UNWIN (Institut für Mathematik, Universität Augsburg, Germany)
Exploring Complex Datasets Graphically with Multiple Views and Interaction
Each complex dataset has to be visualised in different ways, so what is
there that can be said that applies to all?
One strategy is to develop complex graphics to match the complex
datasets.
Graphical displays need to be flexible and interlinked to reveal the broad
range of information that can be found in a dataset.
This presentation discusses interactive graphic tools mixing basic
displays like histograms and scatterplots with more sophisticated
multivariate displays like parallel coordinate plots and mosaic plots.
CHRISTIAN HENNIG (Department of Statistical Science, University College
London)
Asymmetric projection methods in clustering and classification
We discuss linear dimension reduction for data sets which are partitioned
into some known classes in order to separate the
classes as well as possible. The main idea is asymmetric dimension
reduction,
i.e. one group is declared as the “homogeneous” group, which should
appear “packed” in the resulting projection,
while the all other groups should appear separated, though not necessarily
homogeneously, from the homogeneous group.
We will present several version of asymmetry, including some robust
against several types of outliers and
aggregate local information about separation.
PAUL LONGLEY (Department of Geography, University College London)
The quantitative analysis of family names - a contribution to spatial
literacy?
A recent research project has investigated the distribution of surnames in
Great Britain,
both current and historic, in order to understand patterns of regional
economic development, population movement and cultural identity.
This presentation describes the trajectory of interest in the project and
identifies some tentative issues for the
geographic representation of family genealogy.
Tea
FIONN MURTAGH (Department of Computer Science, Royal Holloway, University
of London)
Identifying and exploiting ultrametricity
We begin with pervasive ultrametricity due to high dimensionality and/or
spatial sparsity.
How extent or degree of ultrametricity can be quantified leads us to the
discussion of varied practical cases when
ultrametricity can be partially or locally present in data. We show how
the ultrametricity can be assessed in text or
document collections, in time series signals, and in chemical information
retrieval.
We conclude with a discussion of ultrametricity in astrophysics, relating
to observational cosmology.
SVEN KNUDSEN (Insightful, Ltd)
Statistical Visualisations of Data in the Pharmaceutical Industry
Endless listings and complex tables has been the de-facto standard for
reporting clinical results.
However, recent focus on patient safety has increased both the amount of
data and complexity.
At the same time, reduced budgets are forcing efficiency with which trial
resources are deployed.
As a consequence, both governmental control and the pharmaceutical
industry are starting to embrace
statistical graphs as the process to clear informative realisations of
results and raw data.
This can have several benefits but also surprising consequences, depending
on the stage of the
drug development process. In this talk we focus on the usage of graphs in
the pharmaceutical industry,
in particular what should be included and avoided in a strict regulated
environment.
Examples will be a considerable part of the talk, in particular on single
patient profiles and adverse event review.
Meeting Contact: Dr Mario Cortina-Borja [log in to unmask]
|