|
Post Address : Geoff Nicholls Department of Statistics 1 South Parks Road Oxford, OX1 3TG UK |
Phone: Department +44 1865 281224; College +44 1865 278938 Email: nicholls@stats.ox.ac.uk Office: Peter Medawar building, by the tea room on the 5th floor Maps: [UK/Oxford], [Science area], UK rain, Oxford weather Department: People, Math. genetics + bioinfo and Statistics seminars Reference: catalogue, E-journals, dictionary, minerva and OxCORT |
|
Personalia
I came to Oxford from the Math Department in Auckland, New Zealand in 2005. Here is a picture
of me (in 2003). I have a college homepage at St Peters.
Teaching
The term dates for 2009-2011
and links to homepages for my BS1a Applied Statistics I, DTC modules for
Statistics and ABC and Part A Simulation.
Publications
Scopus lists my publications with citation links
and the ISI maintain a conference calendar.
When the parameter is a partial order - MCMC Bayesian inference
Overheads for ISBA 10/Valencia 9 talk.
Stochastic Dollo models and Language diversification
TraitLab (this link now works) is MatLab software for
phylogenetic tree reconstruction from binary
trait data. It has a manual plus there is a paper and supplement on the theory, with an application to
lexical trait data. See the RSS journal Significance
for a popular article and the arxiv for a recent
technical report from 2009 with DPhil student Robin Ryder. Overheads for my talk at the CRISM
Workshop, Warwick, 30/5/10.
Statistical work on a Spatio-Temporal change-point process
Overheads for seminar in Glasgow 09/12/09.
Prof Patrick Nunn and I have been using a spatial process related to the Richardson cluster model
to make spatial smoothing of temporal change-points in radiocarbon accumulation-rate data. We have
submitted a tech report (2009) with some pages of supplementary material.
Parallel MCMC
Here is a very simple strategy for parallelising a single MCMC run. The idea was used in the physics and
engineering literature before 1990 in the context of simulated annealing. There is a tech. report
in the works.
Downloads
MCMC basics: chapters 7,8 and 9 of my Auckland Physics 707 Inverse Problems lecture
notes (with Fox and Tan). I had some input to the Radiocarbon-dating software (DateLab) which
Martin Jones wrote. It had some nice features (simplicity, rejection sampling as well as MCMC,
and reliable Bayes-factors). I am happy to hear from users.
