School of Mathematical Sciences
Queen Mary, University of London
Autumn 2002
STATISTICS SEMINAR: DESIGN OF EXPERIMENTS
All are welcome
The talks are held at 16.30, all in the Maths Seminar Room (103)
on Level 1 of the Mathematics Building,
Queen Mary, University of London.
The nearest underground station is Stepney Green.
Turn left at the exit and walk 400 yards.
___________________________________________________________
17 October 2002
R. Edmondson, J. Jones
Horticulture Research International
Wellesbourne, Warwick
"Prototype Website for Statistical Design and Analysis of Experiments"
------------------------------------------------------------
31 October 2002
G.M. Arnold
Long Ashton Research Station
Department of Agricultural Sciences
University of Bristol
"Which Willows Do Beetles Prefer?
An Application of 'Square' and 'Circular' Designs"
-------------------------------------------------------------
28 November 2002
S. Pollock
Deparment of Economics
Queen Mary, University of London
"The Design and Application of Frequency-Selective Filters:
A Methodology for Short Nonstationary Sequences"
--------------------------------------------------------------
12 December 2002
R.A. Bailey
School of Mathematical Sciences
Queen Mary, University of London
"Designs for Diallels and Mixtures"
---------------------------------------------------------------
A B S T R A C T S
-----------------
R. Edmondson, J. Jones, Y. Walker
Prototype Website for Statistical Design and Analysis of Experiments
The design of efficient comparative experiments for effective research is
important in many areas of science and technology. Where skilled expert
advice is available it is always desirable to explore and discuss the
various options available at the design stage of an experiment. Ideally,
this should involve a comparison of the various block and treatment designs
that could be used for a particular experiment and the various costs and
benefits of the available options. The expert statistician should discuss
the various options with a client and should make an examination of the
various designs available, perhaps by using simulated data sets.
Unfortunately, for various reasons, this ideal scenario is often
unrealistic. Expert statisticians with the necessary skills may be
unavailable or the client may not have the necessary understanding to
appreciate the various options. Furthermore, the time required to manually
construct and analyse a range of alternative designs may be prohibitive.
We aim to demonstrate a prototype website that is intended to facilitate
and improve the efficient design of experiments by facilitating the
exploration and comparison of different design options at the design stage
of an experiment.
We will aim to present our talk in three sections:
i) An introduction and background to the problem
ii) A 'live' presentation of the construction and analysis of some
of the designs currently available at our prototype website
iii) An outline of the potential for future development
WEBSITE ADDRESS
The site should be available 'live' on the Internet at:
http://www.hri.ac.uk/ExperimentalDesigns/Website/HRI.htm
We would welcome any comments or suggestions for the future development of
the site but please remember that the site is only
a prototype and is still under development.
---------------------------------------------------------------------
G.M. Arnold
Which Willows Do Beetles Prefer?
An Application of 'Square' and 'Circular' Designs
There has been increasing interest in western Europe over the past two
decades in growing short-rotation coppice trees such as willow (Salix spp.)
to provide non-polluting renewable energy sources.
The leaf-feeding chrysomelid beetles are amongst the most common pests of
short-rotation coppice willow. Being a low value crop it is not economic
to use spray programmes of pesticides to control the beetles so it has been
important to identify clones which are more resistant to beetle damage. A
number of different feeding trials have been carried out to attempt both to
quantify likely damage of individual clones and to select new clones with
high resistance. Some background to this research programme and the types
and methods of construction of 'square' and 'circular' designs used for
the feeding preference trials will be presented.
---------------------------------------------------------------------
S. Pollock
The Design and Application of Frequency-Selective Filters:
A Methodology for Short Nonstationary Sequences
This lecture will expound a methodology for trend estimation and
signal extraction that is appropriate to short nonstationary data
sequences. The classical Wiener-Kolmogorov theory of statistical
signal extraction presupposes a data sequence of an indefinite length
and of a stationary nature. The problem of adapting the theory to
cope with tended series of short duration has caused considerable
perplexity; but some simple solutions are offered that appear to be
definite.
The resulting techniques will be illustrated with examples from
meteorology and economics. In meteorology, there is currently much
interest in the problems of discerning the underlying trends in
various global climatic indices such as the index of the annual
temperature anomalies in the global hemispheres or the index of
atmospheric carbon-dioxide concentration.
In the case of the temperature anomalies, we aim to strip away the
noise of the short-term effects and of the measurement errors so as
to clarify the trend. In the case of the carbon dioxide index, we see
a remorselessly upward trend which is surrounded by annual
fluctuations of a remarkable regularity. Here, both the trend and the
annual cycles which surround it are of interest; and we aim to
extract both of these components.
Seasonal effects which are analogous to those present in the carbon
dioxide index can be found in many indices of economic activity.
In some cases, where the components of an aggregate econometric
series reside in disjoint frequency bands, it is possible to use the
filtering techniques to decompose the series into its component parts
and to reveal hidden structure that are concealed within aggregate
series. The Swiss index of national unemployment provides a startling
illustration.
--------------------------------------------------------------------
R.A. Bailey
Designs for Diallels and Mixtures
In some experiments the treatments are effectively the k-subsets
of an n-set. Examples include diallel experiments (k=2); weighing
designs; and mixtures of equal amounts of seed from k species.
If the response can be modelled as the analogue of main effects
(called general combining ability in the diallel case) then the
experiment need not include all subsets. Then it becomes formally
equivalent to an incomplete-block design in which only block totals
are measured. Except in special cases, designs which are optimal for
one situation may be very bad for the other one.
For a more general model, the triangular association scheme (and the
Johnson scheme) seem appropriate. The corresponding partially
balanced designs are generally balanced with respect to the natural
treatment subspaces.
____________________________________________________________________
For more information ask:
Barbara Bogacka
School of Mathematical Sciences
Queen Mary, University of London
Mile End Road
London E1 4NS
Tel: 020 7882 5497
e-mail: [log in to unmask]
The seminar information is kept on:
http://www.maths.qmw.ac.uk/~rab/seminars.html
|