RSSNE Past talks
Royal Statistical Society North Eastern Local Group
Programmes for previous years
Programme 2015/2016
 Friday 13th November : Mike West (Duke University). Revisiting Lindley '92: "Is our view of Bayesian statistics too narrow?"
 Wednesday 16th December : Holly Ainsworth (Newcastle University) + Xmas drinks. Investigating causal pathways between genetic variation and human disease by incorporating "omics" data.
 Tuesday 19th January : Ben Daniels (HMRC) and Weichao Wang (DWP). Practical applications of data science techniques by Government departments. Talk 1: Understanding agents; Talk 2: Analysing DWP complaints data using topic analysis.
 Thursday 3rd March: James Carpenter (London School of Hygiene & Tropical Medicine). Reference based sensitivity analysis for clinical trials with missing data: some theory and examples.
 Friday 18th March: Malcolm Farrow (Newcastle University), Jo Kaczmarska (Risk Management Solutions Inc.) and Francesco Serinaldi (Newcastle University). Challenges in Modelling and Forecasting Rainfall: A Joint Meeting of the Environmental Statistics Section of the RSS and the RSS NE Local Group.
 Tuesday 10th May : Tim Rakow (King's College London). Some experiments on communicating complex statistical data: Helping people understand riskadjusted outcomes data for children's heart surgery.
Programme 2014/2015
 Thursday 18th September : Paul Garthwaite (Open University). Quantifying expert opinion as a probability distribution.
 Wednesday 3rd December : Nancy Singh, Mary Higgs and Penny Sinclair (DWP). Talk 1: Households below average income (HBAI) data collection in the FRS through to the production and use of HBAI poverty and lowincome statistics; Talk 2: An introduction to dynamic microsimulation modelling in DWP.
 Tuesday 16th December : Rachel Oughton (Durham University) + Xmas drinks. Using Bayesian networks for pore pressure prediction.
 Thursday 12th March : Chris Sherlock (Lancaster University). Motor unit number estimation via Bayesian model selection using sequential Monte Carlo.
 Tuesday 28th April: Keith Beven (Lancaster University). Nonstationarity, inconsistencies, heterogeneity, uniqueness of place, and surprises as epistemic uncertainties in hydrological modelling  a framework for good practice.
 Wednesday 20th May : David Banks (Duke University). Bayesian regression models for metabolomic data.
Programme 2013/2014
 Thursday 7th November Gayll Thomson and Adam Pearce (DWP).Data Visualisation
 Thursday 28th November Ian McHale (Salford University). Time varying rankings models to determine the greatest sports players and teams of all time.
 Tuesday 17th December Sarah Heaps (Newcastle University). Bayesian modelling of rainfall data using nonhomogeneous hidden Markov
models and latent Gaussian variables
 Thursday 20th February Jill Johnson (Leeds University). Using statistical emulation to determine parameter sensitivities to the uncertainty in model simulators for cloud and aerosol processes.
 Wednesday 19th March Jenny Barrett (Leeds University) and Philip Li (Northumbria University). Genetic Epidemiology of Melanoma and Using small area data for surveillance and policy evaluation: a Bayesian spatiotemporal approach.
 Tuesday 20th May Sally McClean (University of Ulster). Using Markov and semiMarkov Models to plan Stroke Patient Care
Programme 2012/2013
 Tuesday 2nd October : Kaisey Mandel (Imperial Centre for Inference and Cosmology, Imperial College London). Improving Cosmological Distances to Illuminate Dark Energy:
Hierarchical Bayesian Models for Type Ia Supernovae in the Optical and
NearInfrared.
 Thursday 22nd November : Paul Ainsworth and Mauricio Armellini (DWP). Using Propensity Score Matching to estimate the impact of DWP employment schemes and Estimates of the number of people facing inadequate retirement incomes.
 Tuesday 18th December : Danny Williamson (Durham University) + Xmas drinks. Managing uncertainty in climate change projections.
 Tuesday 19th February : Nelson Kinnersley (Roche Products Ltd). Elicitation of Expert Beliefs in Drug Development.
 Wednesday 20th March: Peter Diggle (Lancaster University). Statistical Modelling for eHealth.
 Tuesday 9th April : Sarah Darby (Oxford University). Radiationrelated Heart Disease in Breast Cancer.
 Tuesday 7th May : Robert G Aykroyd (University of Leeds). Geometric models for object tracking with electrical tomography data.
Programme 2011/2012
 Wednesday 5th October (Newcastle): Paul Baxter (University of Leeds): Statistical epidemiology  the use of observational and audit data in applied health research
 Thursday 3rd November (Durham) : Jonty Rougier (University of Bristol): Emulating the output of large climate simulators
 Wednesday 23rd November (DWP) : Nicky Tarry: Administrative Data and the Census and Gayll Thomson: Data Visualisation.
 Tuesday 13th December (Durham) : Peter Avery (Newcastle University) Screening for Deep Vein Thrombosis; Analysis of a Trial in A&E +AGM + Xmas Drinks
 Thursday 16th February (Durham) : Deborah Ashby (Imperial College London) What role should formal riskbenefit decisionmaking play in the regulation of medicines?
 Thursday 29th March (Newcastle) : Professor R.A. Bailey, (Queen Mary, University of London) From Rothamsted to Northwick Park: designing experiments to avoid bias and reduce variance
 Thursday 19th April (Newcastle) : Pete Philipson (Northumbria University) A Comparison of Sporting Heroes
 Thursday 12th May (Durham) : John Hinde (NUI Galway) Random Effects, Mixtures and NPMLE
Programme 2010/2011
 Wednesday 20th October (World Statistics Day Meeting, Department for Work and Pensions) : Joanna Bulman (DWP) and Judith Correia (DWP), The Life Opportunities Survey and Early estimates for working age inactive benefits
 Tuesday 9th November (Durham) : Paul Gosling (The Food and Environment Agency), Subjective judgements and transparency about uncertainty in policy making.
 Friday 10th December (Newcastle) : Ian Vernon (Durham) + AGM + Xmas drinks!, Visualising the input space of a galaxy formation simulation.
 Thursday 20th January (Newcastle) : Ryan Dunn (Department of Work and Pensions), Analysis of the DWP working age customer base.
 Friday 4th March (Newcastle) : Phil Giles (Smartodds), Applications of Statistics in Sports and Sports Betting.
 Tuesday 5th April (Newcastle) : ONS National Consultation on measuring National Wellbeing.
 Thursday 12th May (Durham) : Volodya Vovk (Royal Holloway), Gametheoretic probability: a brief introduction.
Programme 2009/2010
 Thursday June 3rd 2010 (Durham): Roland Fried (Dortmund), Monitoring medical time series.
 Wednesday March 17th 2010 (Newcastle): Frank Duckworth and Phil Scarfe (Salford), The D/L method in oneday cricket: 13 years on and Decision problems in sport: Naismith's rule and route choice in
adventure racing.
 Tuesday 16th Febraury 2010 (Durham): Lisa Mueller (Sustrans), Does cycling increase?
 Thursday 21st January 2010 (Newcastle): Stephen Bennett (Unilever), "What does the consumer think?  Multivariate statistics for
sensory and consumer analysis in the HPC industry".

 Thursday 15th October 2009 (Newcastle): Nuala Sheehan (Leicester), Inferring causality in observational epidemiology.
 Tuesday 17th November 2009 (Durham): Colin Fyfe (Paisley), using Bregman divergences for exploratory data analysis.
 Thursday 10th December 2009 (Newcastle): Tim Davis (Jaguar and Land Rover) and Tony Greenfield (Newcastle), The rites and prangs of statistical fashions in industry.
Programme 2008/2009
 Thursday 14th May 2009 (Durham): John Kent (Leeds), Procrustes methods in projective shape analysis.
 Thursday 23rd April 2009 (Newcastle): David Firth (Warwick), Getting it right on election night.
 Tuesday 17th March 2009 (Durham): Cathal Walsh (Trinity College Dublin), Bayesian Methods, Health Technology Assessments and Decision Making: A view from the coalface.
 Tuesday 3rd Febraury 2009 (Newcastle): Cam Donaldson (Newcastle), Valuing life and health.
 Thursday 22nd January 2009 (Newcastle): Alexander Donev (Manchester), Optimal Design Theory.
 Tuesday 9th December 2008
Speakers: Jill Johnson (Newcastle) and Graeme Hickey (Durham):
Titles: Spatial Modelling of Extreme Wind and Rain;Estimating Ecotoxicological Hazardous
Concentrations for Environmental Risk Assessment
Venue: Room CM101, Durham
Time:
4pm
Abstracts: (Jill Johnson) The modelling of extreme events from environmental processes such
as wind and rain is of great importance with respect to
structural design problems and global climate change. When
events from more than one process occur jointly, the damaging
effects and problems caused can often be much more severe than
those from just one individual process alone. For a data set
containing daily mean wind speed (m\sec) and rainfall (mm)
records from twentyfive sites across the UK over the period
January 1961  December 1980, the type of extremal dependence
between mean wind speed and rainfall will be characterised using
a statistic known as the "coefficient of tail dependence",
\sigma. The statistic sigma is estimated as the parameter of
the extremal dependence model of Ledford and Tawn (1996). How
this parameter varies across the UK region will be explored.
Initially, a modelling approach that assumes independence
between all sites across the region will be used, and then a
latent spatial process model will be introduced to examine any
spatial dependence that may be present between the site
estimates. As well as quantifying the spatial correlation across
the region for the parameter of the extremal dependence model,
it is also of interest as to how the occurrence of extreme
events for these two variables varies from site to site across
the region. By defining a binary variable representing whether
or not an extreme event has occurred, a binary spatial model is
developed to assess for any spatial correlation in the
occurrence of extreme events for these environmental processes.
Reference: Ledford, A. & Tawn, J.A. (1996) :Statistics for near
independence in multivariate extreme values. Biometrika 83, 169187.)
(Graeme Hickey) Incl.: Peter Craig (Durham University), Andy Hart (CSL, York),
Ben Kefford (RMIT, Australia), Jason Dunlop (NRW, Australia)
Exposure to toxicants, arising from the multibillion dollar
pesticide industry, in an ecological community is a strictly
monitored process; however, the state of the science is
lacking in rigour. The EU Technical Guidance Document (TGD)
currently states that as few as two species from the proposed
ecological community to be affected must be tested with the
toxicant of interest. An arbitrary and antiquated assessment
factor is then applied posthoc to the minimum of this data in
order to postulate a 'safe concentration', called
the Hazardous Concentration (HC).
We discuss two recent, and separate, suggestions (Hickey et
al. 2008a,b) to estimate the HC from the probabilistic
assumptions of species sensitivity distributions  the
community level generalisation of a doseresponse curve. The
first questions the TGD/Ecotoxliterature driven use of upper
onesided underestimate confidence/credible limits as trigger
values for environmental/regulatory action. The second
estimates the HC for salinity in Australia, a rapidly growing
environmental, ecological and economical challenge. The new
method uses a relatively large amount of toxicity data from a
new experimental procedure which produces\A0many (doubly) censored\A0data which hitherto would usually have been
ignored.
Hickey, G. L., Craig, P. S., and Hart, A, 2008. On the application of
loss functions in determining assessment factors for
ecological risk. Ecotoxicol. Environ. Saf. DOI:
10.1016/j.ecoenv.2008.06.004
Hickey, G.L., Kefford, B.J., Dunlop, J.E., Craig, P.S., 2008. Making
species salinity sensitivity distributions reflective of
naturally occurring communities: using rapid testing and
Bayesian statistics. Environ. Toxicol. Chem. In press. DOI:
10.1897/08079.1
Programme 2007/2008
 Monday
22nd October 2007
Speaker: Andris Abakuks (Birkbeck
College):
Title: The synoptic problem and the triplelink
model
Venue: Room CM101, Durham
Time:
3pm
Synopsis: n New Testament studies, the synoptic problem is concerned with hypotheses about the relationships between the synoptic gospels, i.e., the gospels of Matthew, Mark and Luke. The triplelink model represents a family of possible relationships between the synoptic gospels. Counts of the numbers of verbal agreements between the gospels are examined to investigate which of the possible triplelink models appears to give the best fit to the data.  Wednesday
14th November 2007
Speaker: Peter Challenor
(Southampton)
Title: The Probability of Thermohaline
Collapse and Rapid Climate Change
Venue: Herschel Building, HL4.TR3, Newcastle
Time: 4pm
Synopsis:North West Europe has milder winters than Alaska because the North Atlantic carries vast amounts of heat North. This circulation is driven by differences in water density (rather than the wind) and is known as the thermohaline circulation. In some dynamical models of the climate this circulation can be shut down as the concentration of greenhouse gases in the atmosphere increases. If this were to happen it could take place in a few years (a rapid time scale for the climate) and could have severe consequences for the climate of Europe. The climate models we use are large and complex so it is difficult for us to use traditional Monte Carlo techniques to look at uncertainties and to estimate the risk of such a collapse. We therefore use statistical methods developed to deal with complex computer models. These involve building emulators (statistical approximations) for our code outputs as functions of our inputs. We then use these emulators in place of our models in the Monte Carlo calculations. Using a variety of different emission scenarios we estimate the probability that the thermohaline circulation is severely reduced or even collapses.  Thursday 17th January
2008
Speaker: Ian Dryden (Nottingham)
Title:
Shape Analysis and Molecule Matchingh
Venue: Research Beehive 2.21, Newcastle
Time: 4pm
Synopsis:
The statistical analysis of geometrical objects is increasingly important in a wide variety of disciplines, for example in comparing molecules in bioinformatics. Key aspects of shape analysis involve dealing with geometrical invariances and correspondences between parts of objects. Shape data are inherently nonEuclidean, but with care a wide range of practical analyses can be undertaken. We consider Bayesian methodology for comparing two or more steroid molecules, where the labelling correspondence between atoms is unknown. We initially match a pair of molecules, where one molecule is regarded as random and the other fixed. A type of mixture model is proposed for the point set coordinates, and the parameters of the distribution are a labelling matrix (indicating which pairs of points match) and a concentration parameter. An important property of the likelihood is that it is invariant under rotations and translations of the data. Bayesian inference for the parameters is carried out using Markov chain Monte Carlo simulation, and an approximation is considered for speeding up the simulation algorithm. Extensions to multiple molecule alignment are also introduced, and properties of the shape of the steroid molecules are explored in relation to binding activity.
Reference: Dryden, I.L., Hirst, J.D. and Melville, J.L. (2007). Statistical analysis of unlabelled point sets: comparing molecules in chemoinformatics. Biometrics, 63, 237251.  Wednesday 27th February 2008
Speakers:Doug Altman (Oxford) and Sheila Bird (Cambridge)
Title:Selective nonreporting of findings of randomised trials:
review of evidence and impact on systematic reviews (Doug Altman) and
Good design: an essential component of good science (Sheila Bird)
Venue:
Herschel Building LT2, Newcastle
Time: 35.30
Synopsis (Doug Altman): Nonpublication of the findings of some trials
has been recognised as a potential threat to the validity of
metaanalysis. This practice is usually, and paradoxically, called
"publication bias". Recent new evidence has demonstrated that the
selective reporting of trial outcomes within published studies is an
additional threat to validity. Trial reports may include
preferentially those outcomes with statistically significant results
and, in addition, analyses in publications may differ in important
ways from those specified in the trial protocol. I will review the
evidence of bias from nonreporting, with emphasis on selective
outcome reporting. I will consider the possible impact on
metaanalysis, how systematic reviewers can look for evidence of
selective reporting, and what can be done to improve published reports
of randomised trials.
Synopsis (Shelia Bird):Examples of how good design can improve knowledge on risks, whether of
drugsrelated death soon after release from prison or in firstinman
studies; in surveillance, whether of BSE or H5N1 in wild birds; on
costeffectiveness, whether in health or criminal justice; and on
policy effectiveness by use of formal experiments. Using criminal
justice examples, I challenge the notion of 'complex interventions',
and highlight the role of bad design in the 2007 Scottish electoral
fiasco.
 Tuesday
22nd April 2008
Speaker: Trevor Cox (Unilever)
Title: Measuring Plaque and Its Removal from Teeth
Venue: Herschel Building HL4,TR4, Newcastle
Time:
5pm
Synopsis: Plaque on teeth can cause major periodontal
(gum) diseases such as gingivitis and periodontitis. In order to
maintain a healthy mouth, plaque needs to be removed regularly. The
amount of plaque on a tooth is usually assessed subjectively using a
standard index such as the modified Quigley and Hein index. Various
study designs are used to compare the efficacy of dentifrices and
toothbrushes in removing plaque, the crossover trial being very
popular. This talk discusses the various statistical models that are
used and reported to analyse plaque data, but then delves deeper into
the data using tools from the multivariate analysis toolbox, giving
further insight to patterns of plaque and its removal.
 Thursday
8th May 2008
Speaker: Glenn Shafer (Rutgers Business School/Royal Holloway)
Title: Was Jean Ville a statistician?
Venue: CM101, Durham
Time:
4pm
Synopsis: Jean Ville (19101989) was one of the first students to follow the curriculum in probability and statistics set up by Emile Borel at the University of Paris in 1930. He is known for his critique of Richard von Mises's idea of a collective, for his invention of the idea of a martingale, and for contributions to statistical testing, information theory, mathematical economics, and signal theory. Did you ever hear of Jean Ville before? Was he a statistician?
Programme 2006/2007
 Wednesday
18th October 2006
Speaker: Dr James Carpenter (Medical Statistics
Unit, London School of Hygiene and Tropical Medicine)
Title: Sensitivity
analysis after multiple imputation under missing at random: a weighting approach
Venue: "The Buttery", L401, Merz Court, Newcastle University
Time:
3.30 pm (note this earlier time!)
Special Info: Seaonal Refreshments
after the talk!
Synopsis: Multiple Imputation (MI) is now well established
as a flexible, general, method for the analysis of data sets with missing values.
Most implementations assume the missing data are `Missing At Random' (MAR), i.e.
given the observed data, the reason for the missing data does not depend on the
unseen data. However, although this is a helpful and simplifying working assumption,
it is unlikely to be true in practice. Assessing the sensitivity of the analysis
to the MAR assumption is therefore important. However, there is very limited
MI software for this. Further, analysis of a data set with missing values that
are Not Missing At Random (NMAR) is complicated by the need to extend the MAR
imputation model to include a model for the reason for dropout. Here, we propose
a simple alternative. We first impute under MAR and obtain parameter estimates
for each imputed data set. The overall NMAR parameter estimate is a weighted average
of these parameter estimates, where the weights depend on the assumed degree of
departure from MAR. In some settings, this approach gives results that closely
agree with joint modelling as the number of imputations increases. In others,
it provides ballpark estimates of the results of full NMAR modelling, indicating
the extent to which it is necessary and providing a check on its results.
We
illustrate our approach with a small simulation study, and the analysis of data
from a trial of interventions to improve the quality of peer review.
 Wednesday
6th December 2006
Speaker: Professor Heather Cordell, Institute
of Statistical Genetics, Newcastle University
Title:Multiple Imputation
Methods for Genetic Association Analysis of Complex Traits
Venue: Newcastle
Time: 3.30pm (note this earlier time!)
Synopsis:To test for
association between a disease and a set of linked genetic markers, or to estimate
relative risks of disease, several different methods have been developed. Many
methods for family or case/control data require that individuals be genotyped
at the full set of markers and that phaseknown genotypes (i.e. the combination
of haplotypes present in an individual) can be reconstructed. Individuals with
missing data are excluded from the analysis. This can result in an important decrease
in sample size and a loss of information. A possible solution to this problem
is to use missingdata likelihood methods. We propose an alternative approach,
namely the use of multiple imputation. Briefly, this method consists in estimating
from the available data all possible phased genotypes and their respective posterior
probabilities. These posterior probabilities are then used to generate replicate
imputed data sets via a data augmentation algorithm. We performed simulations
to test this approach for case/control and case/parent trio data and we found
that the multiple imputation procedure generally gave unbiased parameter estimates
with correct type 1 error and confidence interval coverage. Multiple imputation
had some advantages over missing data likelihood methods with regards to ease
of use and model flexibility.  Thursday 18th January
2007
Speaker: Dr Baibing Li, Loughborough University
Title:
Tissue classification with gene expression profiles: a clusteringfunctionbased
approach
Venue: Newcastle
Time: 4.00pm
Synopsis:
In recent years, it has become commonplace for researchers to perform cluster
analysis to identify patterns in gene expression data, thanks to the DNA microarray
technology which has now made it possible to investigate thousands of gene expression
data simultaneously. This talk will review the recently developed clustering method,
the clusteringfunctionbased method, and discuss its links to the Fisher linear
discriminant analysis, the modelbased clustering method, and regression analysis.
To illustrate this new methodology, two case studies for clustering gene expression
data are investigated. The results are compared with that obtained using conventional
clustering methods.  Wednesday 7th March 2007
Speaker: Paul Blackwell, Statistics Department, Sheffield University
Title: Environmental Series observed at Uncertain Times
Venue:
CM101, Dept of Mathematics, Durham
Time: 17.00
Synopsis:
In this talk, I will introduce some statistical problems involving "time series"
where the actual times of observations are not precisely known. The main example
that I will discuss is the construction of the radiocarbon calibration curve,
necessary for accurate radiocarbon dating. Recent points on the curve can be obtained
from carbon samples whose ages are known fairly accurately, from the counting
of tree rings; older points have uncertainty in both their radiocarbon levels
and their ages, from measurement or counting error. I will talk about the Bayesian
methods which were developed to tackle this complexity and which led to the latest
international standard curves in 2004. A second example relates to the dating
of measurements from Arctic and Antarctic ice cores. These cores hold crucial
information about the past environment, but interpreting them depends on the complex
and uncertain relationship between the depth of the ice and its age. A new project,
in collaboration with the British Antarctic Survey, involves developing methods
to date ice cores while making due allowance for uncertainty in layer counting,
past rates of ice accumulation and ice flow and compression.
 Thursday
10th May 2007
Speaker: Adrian Bowman, Department of Statistics,
University of Glasgow
Title: The use of additive models in environmental
modelling
Venue: CM101 Durham
Time: 16.30
Synopsis:Additive
models extend standard regression methods by allowing very flexible, but smooth,
relationships between variables of interest. The basic ideas and potential benefits
of these models will be discussed and illustrated on data from several different
environmental studies, including water quality in the River Clyde and SO2 pollution
over Europe. The technical aspects of the talk will focus on appropriate methods
of inference, including the need to incorporate suitable forms of spatial and
temporal correlation.
Programme 2005/2006
 Tuesday 11th October 2005
Speaker:Professor
Christl Donnelly (Dept of Infectious Disease Epidemiology, Imperial College London)
Title: Statistical and epidemiological analysis of the SARS epidemic in
Hong Kong
Venue: L401 (the "Buttery"), Merz Court, Newcastle; contact
Malcolm Farrow, tel 0191
515 2762
Time: 5.30pm
Synopsis: The rapid worldwide spread
of SARS provided a major challenge to public health in 2003. The high fatality
rate and the travel restrictions imposed to control its spread caused widespread
anxiety. Data from Hong Kong will be presented together with the analysis used
to estimate key epidemiological factors determining the spread of SARS and leading
to its control.  Thursday 10 November 2005
Speaker: Professor John Haslett (Dept Statistics, Trinity College Dublin)
Title: Bayesian palaeoclimate reconstruction
Venue: CY91 (previously
CG91), Durham; contact Jonathan Rougier,
tel 0191 334 3111
Time: 5.30pm
Synopsis: We consider the
problem of reconstructing prehistoric climates using fossil data extracted from
lake sediment cores. Such reconstructions promise to provide one of the few ways
to validate modern models of climate change. A hierarchical Bayesian modelling
approach is presented and its use, inversely, is demonstrated in a relatively
small but statistically challenging exercise, the reconstruction of prehistoric
climate at Glendalough in Ireland from fossil pollen. This computationally intensive
method extends current approaches by (a) explicitly modelling uncertainty and
(b) reconstructing entire climate histories. The statistical issues raised relate
to the use of compositional data (pollen) with covariates (climate) which are
available at many modern sites but are missing for the fossil data. The compositional
data arise as mixtures and the missing covariates have a temporal structure. Novel
aspects of the analysis include:  a spatial process model for compositional
data;
 local modelling of lattice data;
 the use, as a prior, of a random
walk with long tailed increments;
 a twostage implementation of the Markov
chain Monte Carlo approach;
 and a fast approximate procedure for crossvalidation
in inverse problems.
We present some details, contrasting its reconstructions
with those generated by a method in use in the palaeoclimatology literature. We
suggest that the method provides a basis for resolving important challenging issues
in palaeoclimate research. We draw attention to several challenging statistical
issues that need to be overcome.  Tuesday 13
December 2005
Speaker: Dr Jonathan Rougier (Dept Mathematical Sciences,
Durham)
Title: Uncertainties concerning future climate
Venue:
CY60 (previously CG60), Durham; contact John
Little, tel 0191 334 3117
Time: 5.30pm
Synopsis: Our
predictions for future climate, and in particular the ways in which it will respond
to increasing levels of atmospheric greenhouse gases, depend to a large extent
on the evaluation of computerbased climate simulators. Statisticians have been
studying this type of inference for many years, under the general heading of Computer
Experiments. Longterm climate prediction presents three particular challenges:
 The simulators are very expensive to evaluate;
 They are generally
quite poor at representing climate at the regional level (which is where most
of the interest resides);
 Lots of interesting things could happen in the next
100 years (eg in technology, economics and demographics).
We look at the
ways in which these problems contribute to our uncertainty about future climate.
This is the traditional Christmas meeting, to be accompanied by cheese ("cheez"
for vegans) and wine.
 Tuesday 7 February 2006
Speaker: Dr William Browne (School of Mathematical Sciences, Univ. of Nottingham)
Title: Classification of mass spectroscopy data using principal components
analysis, Bayesian MCMC modelling and a deterministic peak finding algorithm.
Venue: L401 (the "Buttery"), Merz Court, Newcastle; contact Darren
Wilkinson, tel 0191 222 7320
Time: 5.30pm
Synopsis:
In this talk we consider three approaches to classifying SELDI and MALDI mass
spectroscopy datasets. Individual mass spectroscopy scans consist of a trace of
~14,000 values (intensities) at differing mass to charge ratios. Each individual
scan belongs to one of m groups where individual groups may represent differing
breast cancer lines and we wish to classify new scans to these groups. Due to
the large number of variables (mass/charge ratios) associated with each scan we
require data reduction techniques to give a group of derived variables that can
be used for classification. We consider three techniques: Firstly principal components
analysis (PCA) of the full scans to produce a smaller group of derived variables.
Then two methods that take into account the fact that the scan is a sequential
set of variables, and attempt to fit mixtures of scaled Gaussian distribution
functions to the scan. We consider a deterministic algorithm and a model based
MCMC method that reduce each scan to a series of (scaled) Gaussian peaks at locations
that are common to all scans. The resulting heights of these peaks are then used
in the classification. All three methods will be compared via crossvalidation
on two example datasets, one with 6 groups and one with 2 groups. This is joint
work with Ian Dryden and Kelly Handley.  Thursday
9 March 2006
Speaker: Dr Barbel Finkenstadt (Department of Statistics,
University of Warwick)
Title: Stochastic modelling and inference for
transcriptional gene regulation
Venue: L401 (the "Buttery"), Merz Court,
Newcastle; contact Malcolm Farrow,
tel 0191 515 2762
Time: 5.30pm
Synopsis: A stochastic distributed
delayed model is developed that explains the general molecular process of gene
regulation. The model incorporates gene transcription, translation into protein
and repression of gene expression by the nuclear form of the protein. Markov chain
Monte Carlo methods are used to make inference about these unobserved processes
and the unknown parameters of interest. We illustrate the methodology using a
well understood gene regulatory network, namely the circadian clock, as an example.
This is a joint meeting between the RSSNE Group and BioNEt. Registration will be required;
this can be done at the meeting itself. Register by emailing Natasha
Taylor with name/affiliation.
 Thursday
18 May 2006
Speaker: Professor Colin Aitken (Edinburgh University)
Title: Evaluation of evidence
Venue: CY60 (Chemistry Dept),
Durham; contact Frank Coolen, tel
0191 334 3048
Time: 5.30pm
Synopsis: A review of probabilistic
approaches to the evaluation of evidence will be given. This will include relative
frequencies, significance probabilities, and likelihood ratios. Evidence evaluation
for multivariate, twolevel models, where the betweengroup distribution is not
assumed to be normally distributed, will be discussed with the aid of graphical
models. Reference will be made to some uses of statistical ideas in the courts.
Cases discussed will include (a) the Daubert, Joiner and Kuhmo Tire
trilogy from the US Supreme Court where the requirements for the admissibility
of expert evidence were expressed, (b) the English Court of Appeal cases R
v. Adams, D.J. and R. v. Doheny and Adams where comments about the
role of statistics in legal argument were made, and (c) a more recent English
Court of Appeal case R v. Grey where comments were made about the requirements
for the use of frequency data. The relevance of these for recent cases concerning
sudden explained deaths of infants will be discussed.
 Daubert v. Merrell
Dow Pharmaceuticals Inc., 509 U.S. 579,1993.
 General Electric Co. v Joiner,
522 U.S. 136, 1997.
 Kumho Tire Co. Ltd. v. Carmichael, 526 U.S. 137, 1999.
 R v. Adams, D.J., 1997, 2 Cr. App. Rep. 4679.
 R v. Doheny and Adams, 1
Cr. App. R. 369, 375, 1997.
 R. v. Grey, CA(Crim Div), [2003] EWCA Crim 1001.
Programme 2004/2005
Programme
2003/2004
 Thursday 11th December 2003
Speaker: Jonathan Crook (School of Management, University of Edinburgh)
Title: Reject inference techniques in credit scoring
Venue:
Room L401, Merz Court, University of Newcastle upon Tyne
Time: 5.30
pm
Synopsis: The estimated parameters of application scorecards may
not be accurate estimates of the parameters that would relate to the population
of all applicants because rejected applicants are usually excluded from the sample
used to estimate these parameters. This work employs a sample that includes those
who would normally be rejected in order to examine the extent to which the exclusion
of rejected applicants undermines the predictive performance of a scorecard based
only on accepted applicants and the extent to which reject inference techniques
can remedy the influence of his exclusion. This analysis is applied to five acceptance
thresholds. We also discuss the performance of the bivariate probit model with
selection as an alternative reject inference technique.  Thursday
22nd January 2004
Speaker: Byron Morgan (University of Kent)
Title: Models for yeast prions
Venue: Room L401, Merz Court,
University of Newcastle upon Tyne
Time: 5.30 pm
Synopsis:
Yeast cells may contain proteins called prions, that behave like the mammalian
prion PrP, implicated in CJD. This talk describes multitype branching process
modelling of the reproduction of yeast cells, and the distribution of prions in
the cell population. The models are being developed in conjunction with current
laboratory experiments.  Thursday 11th March
2004
Speaker: Brian Ripley (University of Oxford)
Title:
Statistical Methods need Software
Venue: Room CG232, Science
Site, University of Durham
Time: 5.30 pm
Synopsis: There
is a chasm between writing a paper on a new method of data analysis for a journal
and actually being able to use it for a real applied problem (and that may explain
why the papers so rarely do supply such examples). For people trying to teach
modern applied statistics the connection between theory and practice is vital.
The missing bridge is suitable software, and the talk will cover some of the work
I and collaborators have done to try to supply such a bridge (in particular the
R project) and end with some examples of data analyses enabled by such software.
 Friday 23rd April 2004
Speaker:
Doug Montgomery (Arizona State University)
Title: Designing Experiments:
Some Adventures and Lessons Learned
Venue: Continuing Professional
Development room, Stephenson Centre, second floor Stephenson Building, University
of Newcastle upon Tyne
Time: 3.30 pm
Synopsis: The most
important phases of any experiment are preexperimental planning and actual test
conduct. If these activities are carried out effectively, then there is usually
a high probability of success. However, planning and execution of an experiment
present many opportunities for things to go wrong. This presentation illustrates
several of these opportunities, based on the speaker's experiences in both academia
and industry. Some of these experiences have resulted in (sometimes painful) learning
opportunities. Suggestions for avoiding these problems are given.  Friday
28th May 2004
Speaker: Stephen Senn (University of Glasgow)
Title: Propensity Score or Analysis of Covariance?
Venue: Room
L401, Merz Court, University of Newcastle upon Tyne
Time: 5.30 pm
Synopsis: An increasingly popular approach to adjusting for possible confounding
in epidemiological studies and clinical trials is to stratify by the socalled
'propensity score'. I consider some disadvantages of this compared to more traditional
approaches such as that of analysis of covariance.
Programme 2002/2003
 Thursday 31st
October 2002
Speaker: John Marriott (Nottingham Trent University
)
Title: A Bayesian approach to estimating smooth trends in time series
Venue: Room CG60, University of Durham
Time: 5.30 pm
Synopsis: Many real time series exhibit changes
in trend. However, because the observed changes are frequently not abrupt, models
that specify step changes do not always adequately represent the observed behaviour.
One approach to the problem has been to use smooth transition models that allow
for a gradual change from one linear regime to another. This talk is concerned
with a Bayesian approach to parameter estimation and prediction for smooth transition
models.
 Thursday 21st November 2002
Speaker: Simon Godsill (University of Cambridge)
Title: From
restoration of gramophone recordings to musical sound analysis: a collection of
Bayesian methods
Venue: Room CG60, University
of Durham
Time: 5.30 pm
 Thursday
12th December 2002
Speaker: Robert Aykroyd (University of Leeds)
Title: Bayesian modelling for industrial process tomography
Venue:
Room L401, Merz Court, University
of Newcastle upon Tyne
Time: 5.30 pm
Synopsis: Electrical
impedance tomography is a noninvasive technique used to visualise processes within
physically inaccessible areas by taking measurements around their boundary. There
are many potential applications to industrial processes for online monitoring
and control of evolving processes. Reconstruction of images is a nonlinear
inverse problem. To yield a stable solution it is necessary to make considerable
use of prior information. The role of a Bayesian approach is therefore of fundamental
importance, especially when coupled with MCMC sampling to provide information
about solution behaviour.
 Thursday 6th
February 2003
Speaker: David Spiegelhalter (MRC Biostatistics Unit,
Cambridge)
Title: Monitoring medical outcomes: Bristol, Shipman and
beyond
Venue: Demonstration Lecture Theatre, 4th Floor Cookson Building,
Medical School, University of Newcastle
upon Tyne
Time: 5.30 pm
Synopsis: We show how statistical
methods with a background in quality control can be adapted to a clinical context.
These include funnel plots, riskadjusted CUSUMS, and riskadjusted sequential
probability ratio tests. In particular we see what might have happened had such
techniques been prospectively applied to the Bristol and Shipman cases. 
Thursday, 20th March 2003
Speaker: David
Howard, (Dept of Speech, University of Newcastle upon Tyne)
Title:
Statistics in practice: single cases and case series in neuropsychology
Venue:
Buttery (L401), 4th Floor Merz Court, University
of Newcastle upon Tyne
Time: 5.30 pm
Synopsis: Cognitive
neuropsychology typically uses accuracy data from single cases to investigate
cognitive functioning. I will briefly discuss the reasons behind this approach
and describe current statistical practice and malpractice in the field and some
of the reasons behind this. Then I will consider some problems where ready made
statistical solutions do not exist, and suggest, for discussion, some ways in
which these problems may be addressed.  Thursday
8th May 2003
Speaker: Ernst Wit (University of Glasgow )
Title:
Can we make statistics count in bioinformatics?
Venue: Room CG60,
University of Durham
Time: 5.30 pm
Synopsis: Statistics
should be the science par excellence to be leading bioinformatical endeavours.
The appreciation of uncertainty and variation as a general feature of bioinformatical
data typifies the enormous contribution it can make. In particular, principles
of sound design are required to get workable data; exploratory methods can be
used for assessing data quality and for getting preliminary results; formal highdimensional
methods coupled with statistical awareness of variation yield estimates and predictions
together with measures of reliability. In this talk, I'll explore the current
practice of bioinformatics and the role of statistics within it. I'll present
examples of what statistics can contribute and what it has to offer for the future.
Programme 2001/2002
 Thursday
15th November 2001
Speaker: Deborah Ashby (Queen Mary and Westfield
College, London)
Title: Evidencebased medicine as Bayesian decisionmaking
Venue: Women's Club Room, Senior Common Room, University of Newcastle
upon Tyne.
Time: 5.30 pm
Synopsis: Evidencebased medicine
requires an integrated assessment of available evidence, with its associated uncertainty.
There is also an emphasis on decisionmaking, demanding consideration of values
and costs (utilities) associated with potential outcomes. We argue that the natural
statistical framework is Bayesian decisionmaking, and propose a practical agenda
for further development.  Tuesday 11th December,
2001
Speaker: Darren Wilkinson (University of Newcastle upon Tyne)
Title: Stochastic modelling and inference for genetic regulatory networks
Venue: Lecture Theatre 4, King George VI Building, University of Newcastle
upon Tyne.
Time: 5.30 pm
Synopsis: It is generally acknowledged
that the molecular mechanisms regulating key cellular processes such as gene expression
are intrinsically stochastic. The random diffusion of cell signalling molecules
and the combinatorial assembly of transcription factor complexes provide extensive
opportunities for the action of chance. In recent years stochastic regulatory
network models have been developed, based on discreteevent simulation techniques
for generating realisations from the complex continuoustime countablestate Markov
processes governing the reaction systems. These models contain many parameters
with uncertain values. In addition, the latent process can only be observed partially,
and at discrete time intervals. Inference for such Markov process models is an
extremely challenging problem. This talk will describe the techniques used
to model regulatory networks, and the computational tools needed for simulation
and analysis. An overview will also be given of the MCMC algorithms which can
in principle be used for carrying out Bayesian inference for the parameters underlying
the network models, and the problems associated with applying such techniques
in practice.
The
project's web page gives more details.
 Thursday
17th January, 2002
Speaker: Chris Brunsdon (Department of Geography,
University of Newcastle upon Tyne)
Title: Using Shape Space Analysis
To Explore Map Patterns
Venue: Room Ph30, Department of Physics, University
of Durham.
Time: 5.30 pm
Synopsis: There are a surprising
number of statistical techniques that are of interest to geographers. Shape space
analysis is one technique that seems to hold promise. Here the technique is applied
to two problems: one related to identifying authorship and lineage of antique
maps, and another considering the growth patterns of urban areas. Emphasis will
be placed on graphical and exploratory methods, although some formal hypothesis
testing will be presented.  Thursday 21st February,
2002
Speaker: Henry Wynn (University of Warwick)
Title:
Bayes nets and time series: trying to build causal diagrams in multivariate SPC
Venue: Room G11, Percy Building, University of Newcastle upon Tyne
Time: 5.30 pm
Last modified: Mon Aug 22 15:48:46 GMT 2016
>