PhD: Designs for spatially or temporally correlated data
Liverpool John Moores University and IACR-Rothamsted
There have been considerable advances recently in methods, algorithms and
software for the analysis of correlated data. These offer exciting
possibilities for extending the conventional methods of analysis, to cater for
continuous trends of variability in addition to the discrete "blocking" methods
employed in classical analysis of variance. They are particularly useful in
studies where the modelling of the spatial variation is main interest as may be
the case, for example, in many environmental studies. Furthermore, the
variation can equally well be over time as well as, or instead of, in space as
occurs in many medical and ecological studies.
To exploit these analysis techniques fully, new types of design need to be
developed to enable the random trends to be estimated most effectively.
Initially the student will review the existing methodology to assess the
strengths and weaknesses of the currently available techniques. The student
will then research into new algorithms, in particular, to cater for more than
the two dimensions that are usually considered. Algorithms will also be
investigated to generate designs for multi-dimensional non linear response
surface models with continuous treatments, in addition to the more usual
discrete treatment factors. A key objective of the research will be to
establish methods that produce designs that are robust to a wide range of types
of correlation.
The PhD will be supervised by Professor Roger Payne (Rothamsted and Liverpool
John Moores) and Dr Simon Kirby (Liverpool John Moores). It will be based
mainly at Liverpool but with time spent also at Rothamsted for practical
projects. The student will receive a minimum grant of 6130 pounds a year and
there is the possibility of earning further money by undertaking some teaching
activities.
For further information, please contact Roger Payne, Statistics Department,
IACR-Rothamsted, Harpenden, Herts AL5 2JQ (e-mail [log in to unmask])
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|