The following seminars will take place at Goldsmiths College on Friday
21st May.
2.00 pm Martin Newby (City University)
Analysis of degradation models.
3.30 pm Paul Blackwell (University of Sheffield)
Bayesian inference for random tessellation models.
The seminars will take place in room 137 in the main college building.
The college is a short walk from New Cross and New Cross Gate Rail
and Underground Stations.
Abstracts:
Analysis of degradation models
The work is prompted by the need for a probabilistic analysis of fatigue
crack growth in engineering design, and by the need to devise monitoring
schemes for fatigue failure in structures.
A standard model for fatigue crack growth, The Paris-Erdogan Model, has
a stochastic process counterpart which can be transformed to a Brownian
motion representation of the process. The transformation is the Box-Cox
power transformation.
The analysis exploits the well known observation that the likelihood for
a Wiener process is equivalent to that for a linear model which together
with the Box-Cox transformation allows a Bayesian analysis of several
sets of fatigue crack growth data.
Bayesian inference for random tessellation models
Random tessellations constitute an important class of models in many areas
of application, including ecology, geography and, on a much smaller
physical scale, the structure of materials. Given such a tessellation, an
inhomogeneous Poisson process can be defined, in which the intensity at a
point depends only on its location relative to the tessellation.
I will describe an approach that enables fully Bayesian inference for both
the realisation of the tessellation process and the parameters of the
derived point process, using a random walk Markov chain Monte Carlo
algorithm.
I will illustrate the method with an application in ecology, in which the
tiles of the tessellation represent a set of territories of badger (Meles
meles) clans, and the point process models the occurrence of latrines,
which mark boundaries.
Lawrence Pettit
Dept of Mathematical and Computing Sciences
Goldsmiths College
e-mail: [log in to unmask]
tel: 0171 919 7863
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|