WORKSHOP ON STATISTICAL ANALYSIS OF
COMPUTER CODE OUTPUTS
CALL FOR EXPRESSIONS OF INTEREST
A workshop on "Statistical Analysis of Computer Code Outputs",
supported by EPSRC, will be held on the 10th to 14th April,
2000, at Gregynog in Mid-Wales.
** Topic: The title of the workshop refers to statistical methods
for tackling problems in the use of complex computer models. If
we regard the computer code as a function f relating the inputs x to
the outputs y through y=f(x), we are concerned in general with
various problems of inference about the outputs.
Specific inferences include:-
Interpolation - inference about f(x) at some unobserved input
configuration x.
Sensitivity analysis - inference about how f responds to changes in
one or more input parameters.
Uncertainty analysis - inference about the distribution of the
random output Y=f(X) induced by some distribution for a random
input X.
Calibration - inference about the "true" values of one or more
unknown input parameters appropriate to a particular physical
application of the computer model.
There are various ways of tackling these problems by
statistical methods. For example, some methods rely on a
Monte Carlo approach which runs the code to evaluate the
output at a random sample of input configurations. Such
methods typically demand large numbers of runs
of the code, which is impractical in many situations where a single
run of the code can take several minutes, or even hours.
Some approaches treat the code as a black box defined by an
*unknown* function f. Although f is of course known in principle, in
practice we do not know its value f(x) at any given input
configuration x without actually running the computer code. These
methods in effect make inference about f, or about
appropriate functionals of f, using as data the outputs y_1=f(x_1),
..., y_n=f(x_n) from a set of n runs of the code.
This is a relatively new field, and the purpose of the workshop is to
identify and contrast the range of methods currently being employed.
** Organising committee: A O'Hagan (Sheffield), S French (Manchester)
and J Q Smith (Warwick).
** Participation: In addition to the organisers, the following
experts in the field have already expressed their willingness
to take part -
K Beven (Lancaster)
R M Cooke (Delft)
D Draper (Bath)
M Goldstein (Durham)
J Helton (Arizona State)
A E Raftery (Seattle)
D Rios-Insua (Madrid)
J Sacks (Illinois)
A Saltelli (ISPRA, Italy)
E M Scott (Glasgow)
H P Wynn (Warwick)
The organisers are now calling for expressions of interest from
others who wish to participate. The following categories (and
indicative numbers of places in each category) of participants
are sought:
Other established researchers in this or related fields (6, from
any country)
Younger researchers working or interested in working in this
field (10, predominantly from UK, with some preference for those
having less than five years post-doctoral experience)
Research students working or interested in working in this
field (5, from UK only)
If you wish to take part, please email Tony O'Hagan with brief
details of your experience and the reasons for your interest,
** by Friday 30 July, 1999.**
=> [log in to unmask]
** Support: All local expenses will be paid, plus a contribution
towards travel expenses.
** Venue: Gregynog is a country house owned by the University
of Wales. The house itself is of some historical interest, and
is set in beautiful grounds. This is a very peaceful part of
Britain, with good walking and no discoteques or nightclubs!
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|