Wednesday 10th May 2000
Joint Meeting with St Andrews University (in Lecture Theatre B of the
Mathematical Institute at St Andrews University)

Christine Hackett (BIOSS)

Statistical methods for linkage analysis and QTL mapping in plants

The association between genetics and statistics is long-standing and many important statistical concepts were developed, for example by Galton, Pearson and above all Fisher, in response to questionsmotivated by genetics. There were many important developments in statistical genetics in the period 1908-1940's, but analyses from this period tend to involve characters affected by a single gene. Because there are a relatively small number of such characters, practical analysis was limited. In the last 25 years, developments in molecular biology mean that variation can be observed in the DNA of an organism, giving a virtually unlimited supply of molecular markers whose inheritance can be followed.

One use of molecular markers is to develop a linkage map of a species, with positions along each chromosome labelled by molecular markers. Such a map enables the geneticist to locate genes controlling important traits relative to the markers, particularly quantitative traits i.e. those controlled by a large number of genes, and affected by the environment. For quantitative traits, e.g yield or height, a continuous response is observed, and the effects of the individual genes (referred to as quantitative trait loci or QTLs) cannot be observed directly. Mapping studies have been performed in man, domestic animals and agricultural and forest crops. The ease with which experimental crosses can be made and large numbers of offspring raised mean that agricultural crops are the simplest subjects for mapping studies.

This talk will review the modelling of recombination between molecular markers, which forms the basis for estimating a linkage map. From there, statistical methods for QTL mapping will be discussed. These are based on mixture models, and thresholds for significance testing need to considered carefully. Some current work on QTL analysis for multiple traits and multiple environments will be described.

4.30pm: Tea


David Balding (Reading)

Genealogical Modelling to Locate Disease Genes from Case-Control Data

Many methods for locating disease genes are based on genetic and phenotypic data from extended families affected by the disease. However, these are limited in their accuracy by the small number of recombination events underlying even a very large family tree. In contrast, the genealogical tree underlying a population sample of case chromosomes offers many more recombinations, but the problem now is that the tree is unknown. This is problematic because the case chromosomes are not mutually independent; for example, some but not others will have been affected by particular mutation or recombination events in the past. Drawing valid inferences from these data about disease gene location requires an assessment of the patterns of dependence in the data due to sharing of ancestry, which in turn requires accounting for the effects of the underlying genealogical tree. Much of the recent literature simply ignores the dependence problem, leading to over-optimistic assessments of uncertainty in the inferred location. Recently some authors have offered approximations to take the dependence into account. We propose explicitly modelling the genealogical tree underlying the sample of case chromosomes, within the coalescent modelling framework. We implement Bayesian inference under the model via Markov chain Monte Carlo. This is joint work with Andrew Morris and John Whittaker, and is funded by Pfizer UK.

Wednesday June 7th 2000

AGM and committee meeting
3pm room 347 Meston Building University of Aberdeen

The AGM will precede the talk at 5.45pm
6pm room 302 Meston Building King's College University of Aberdeen

Mr Roger Black (Head of the Scottish Cancer Intelligence Unit)

Developments in Health Statistics in Scotland

The 1990s has seen unprecedented demand for statistical information on health and health services in Scotland. This has been mainly due to:

* the Scottish Parliament
* public and media concern about the effectiveness of services such as screening programmes
* an acceptance by clinicians that involvement in local audit, clinical trials and other work in the field of clinical effectiveness tends to improve services
* a requirement by managers to inform decision making in the context of major organisational change, such as the formation of NHS Trusts

This presentation will describe the evolution of the Information and Statistics Division of the NHS in Scotland in response to these challenges, drawing on examples from the field of cancer. Current developments, including the impact of the White Paper on National Statistics, will also be discussed.

4pm Wednesday September 13th 2000

Macauley Land Use Research Institute,

David Fletcher, University of Otago, New Zealand 

 Assessing the Risk to Hector's Dolphin in New Zealand: The Statistical Challenges 

tea at 3.45pm 

6pm,  25 October 2000

Meston Building 302, King's College, University of Aberdeen 

Marion Campbell, Programme Director, Scottish Health Services Research Unit 

 Statistical Applications in Health Services Research 

The Scottish Health Services Research Unit has a national remit to research the best ways to provide health care, and to train those working in the health services in research methods. Most research projects aim to find out whether novel developments within the health service really are effective and cost-effective. Principal methodologies involved in the conduct of these projects are systematic reviews, pragmatic randomised controlled trials, quasi-experimental studies, sample surveys and qualitative methods. In this talk, I will give a brief overview of the statistical methods more commonly used in this field and will highlight a number of these in more detail. In particular, I will summarise work being developed within the Unit around the design and analysis of cluster randomised trials, the incorporation of the þlearning curveü in the evaluation of emerging health technologies and the individual patient data meta-analysis of surgical trials.

tea at 5.45pm

6pm 23rd November 2000
Meston Building 302, Kings College, University of Aberdeen

William Browne, Institute of Education, London

Multilevel modelling in MLwiN : What's new and what's still do come

This talk will be part talk and part software demonstration as I intend to convey some of the basics of multilevel modelling by using the software package MLwiN. My talk will be in two parts. A new version of MLwiN was released to the user community in February this year and contains many new features. In the first part of my talk I intend through the tutorial exam dataset example to explain some of the basics of multilevel modelling, contrast some of the estimation methods and show some of the new graphing and diagnostics procedures. The Multilevel modelling team at the IOE which I am a member of has recently received a large 3 year grant to investigate methods for fitting more complex multilevel models to social science datasets. This constitutes assessing what models social science researchers are likely to want to fit and which method is best for which model. In the second part of my talk (depending on time) I will talk on up to four areas in which I have recently implemented MCMC based solutions to these problems

a) Multilevel models with complex level 1 variation.
b) Cross-classified and multiple membership models.
c) Multivariate normal response multilevel models with missing responses.
d) Measurement error problems in multilevel models.

tea: 5.45pm

6pm, January 30th 2001

Meston Building 302 University of Aberdeen

Rob Fryer (FRS Marine Laboratory Aberdeen)

Desperately seeking superbeeste, the ultimate monitoring organism

Fish are often used to monitor changes in contaminant levels in the marine environment. But are they any good at it? A simple uptake and excretion model takes us into murky waters in search of our fishy grail. And when nature proves elusive, a smoother based model is used to assess time series in which contaminant levels vary with fish size, and the contaminant-size relationship evolves over time.

Tea at :5.45

6pm, February 22nd 2001

Meston Building 302 University of Aberdeen

David J. Hand (Imperial College)

Data mining for fun and profit

Data mining is defined as the process of seeking interesting or valuable information within large data sets. Novel features of the discipline, which distinguish it from statistics and other areas are noted, and the distinction between model building and pattern detection is explored. The fundamental issue of data quality is examined, and the talk is illustrated with a variety of real examples.

Tea at 5.45pm

4pm, Monday 23 April 2001

WB20, Postgraduate Centre, Medical School, University of Aberdeen.

Diana Elbourne.

Medical Statistics Unit, London School of Hygiene
and Tropical Medicine, Keppel Street, London WC1E 7HT, UK

Meta-Analyses Involving Cross-Over Trials: Some Statistical Considerations

Most meta-analyses synthesise evidence from parallel group trials in which individual patients are randomised to two different treatments. Evidence can also be available from trials of different designs. In one such design the sequence of treatments is randomised - crossover trials. In principle, a meta-analysis can be conducted if, for each trial potentially in the meta-analysis, the relevant estimates of treatment effect with appropriate standard errors are provided. If these are available for a cross-over trial, it can usually be included in a meta-analysis, and calculations performed using standard statistical packages. The problems commonly arise when the relevant estimates are not given. If the information cannot be retrieved from the authors, there are a number of options to consider: exclude the trial results from the quantitative synthesis, and refer to it separately; treat the data as if they come from a parallel group trial (This is usually conservative, but can be misleading if there are carryover or period effects); take data from the first period only (This clearly reduces the potential amount of information. Also, the data may either not be available, or only available for a biased subset of trials with carryover effect); or impute estimates of treatment effect and standard errors. Based on work in progress with Francois Curtin, Doug Altman and Julian Higgins, methods will be described separately, depending on whether the data are continuous or binary, and whether or not there is any carryover. Although none of the possibilities is ideal nor likely to be successful in all cases, the approach may have general implications for strategies in meta-analyses involving non-standard designs.

Tea from 3.30pm in the Health Services Research Unit library

Thursday 25th May 2001
Joint Meeting with St Andrews University

(in Lecture Theatre B of the Mathematical Institute at St Andrews University)


Professor Ian Diamond (University of Southampton)

"The Development of a One Number Census in the UK"

The 2001 censuses in the UK will, for the first time, include a strategy to control for underenumeration, which will permit estimates of underenumeration at all levels of aggregation. The development of this strategy has been the result of an integrated programme of research between all three UK censuses. The paper will describe the esign and estimation issues which have been addressed as well as the strategy to be used to quality assure the final results.

4.30pm: Tea


Mr Ian Máté (General Register Office for Scotland)

Provisional title:- "The Census Coverage Survey and One Number Census in Scotland"

Tuesday 5th June 2001

5.45 pm AGM

6pm, 302 Meston Building University of Aberdeen

Mike Titterington

University of Glasgow
Parameter Estimation for Complex Models

The talk will consider difficulties that arise in the analysis of incomplete data. In principle, maximum likelihood or Bayesian inference can be carried out, using tools such as the EM algorithm and Gibbs sampling, but even these can become unwieldy or sometimes impractical. Particular emphasis will laid on labour-saving approximations, such as the variational approximations that have recently been described in the neural-computing literature. Models that will be discussed include simple mixture models and the hidden Markov random field models used in the statistical analysis of noisy images.

Thursday 27th September 2001

6pm, 302 Meston Building University of Aberdeen

Ian Jolliffe

University of Aberdeen

Some statistical aspects of climate variability

Climate variability occurs at all time scales, ranging from millions of years to days. The latter gives us weather, but such variations also contribute to climate variability. A major reason for studying climate variability is in order to decide whether recent changes are compatible with long-term þnaturalü variability, or is caused, at least in part, by human interventions. Statistics clearly has a role to play in this problem, and the subject is also relevant to the study of climate variability in a number of other ways. These include

Some of the topics listed are vast. The talk will give an overview of some of them, providing a flavour of the many ways in which statisticians and climatologists can collaborate on interesting problems.

Thursday 25th October 2001

6pm M302, Postgraduate Centre, Meston Building University of Aberdeen.

Guy Nason

Department of Mathematics, University of Bristol

Wavelets and texture*

The perceived texture of an image depends on the scale that it is observed. In this talk we show how wavelet processes can be used to model and analyse texture. Our wavelet texture models permit the classification of images based on texture and reveal important information on texture differences in scale-location-direction. We provide examples, taken from industry, where wavelet methods have enhanced the classification of images of hair and fabrics.
(*) Joint work with Idris Eckley.

Thursday 22th November 2001

4 PM WB20, Postgraduate Centre, Medical School, University of Aberdeen.

V.T. Farewell

MRC Biostatistics Unit, Cambridge

Monitoring of Surgical Outcomes: An Application of Cumulative Sum Procedures

Process monitoring has been extensively studied in the industrial context and a widely used tool is the cumulative sum (CUSUM) chart. The application of this methodology to the monitoring of binary surgical outcomes will be discussed. Suggestions will be made for the extension of this methodology to deal with bivariate outcome data and the need to adjust for the case mix in the monitoring of individual surgeons. A discussion of alternative monitoring procedures will be given.

Thursday 10th January 2002

6 pm Room 302, Meston Building, University of Aberdeen.

P. E. Jupp

School and Mathematics and Statistics, St Andrews University

Some Applications of Directional Statistics to Astronomy

In many areas of astronomy, information about distance is not available or is unreliable, so that observations reduce to directions (unit vectors) on the celestial sphere. Thus the tools of directional statistics are required. This talk aims to illustrate the special flavour of directional statistics by considering three problems in astronomy, concerning (i) asteroids, (ii) binary stars, (iii) cometary orbits. A complicating factor in all three problems is the presence of selection effects. Two of the problems involve the further difficulty of the "mirror-image ambiguity".

DATE: 4pm, Thursday 14 November 2002 VENUE: 7th floor conference room of the phase 2 part of the Institute of Medical Sciences (IMS) building at the medical school, University of Aberdeen. Phase 2 is under the 1 in and to the right of the B in The IMS building is a very modern 7 storey building of matt grey panels with chrome chimneys. TEA: from 3:45 pm (in the same room) 6pm, Thursday 5 June 2003
6 pm

John Crawford

Thursday 21st February 2002

SIMBIOS University of Abertay Dundee

Why it takes all kinds - linking diversity to function in complex biological communities.

Room 302,
Meston Building,
University of Aberdeen.
The talk will address the issue of complexity in diverse communities, and in particular how we might define complexity in a way that tells us something about the dynamics and functioning of the system. Any way forward must embrace the full extent of diversity between and within species groups (where these can be usefully defined), and resulting theory must incorporate measurable variables. A theoretical framework that aims to tackle some of these challenges will be presented and illustrated through application to a species rich grassland and the soil-microbial community.


Peter S. Craig

Monday 25th March 2002

University of Durham

Statistical Analysis of Computer Code Output

Grampian Room
Macaulay Institute
Many disciplines rely on deterministic computer simulation to understand and forecast reality. The talk will describe some applications, including hydrocarbon reservoirs and pollutant transport on contaminated land, state some standard problems and outline the variety of formal and informal statistical methods which are available for such situations.

Joint St Andrews/Highland RSS Group Meeting on Statistical Analysis of Weather and Climate


Dr. David Stephenson

Wednesday May 1, 2002

University of Reading

Does the Weather God Play Dice?

Lecture Theatre D
School of Mathematics and Statistics,
University of St Andrews
We live inside a thin envelope of fluid - the atmosphere - that exhibits fascinating variations on all time and space scales. Daily "weather" may be considered to be a realisation of a random process, whose "long-term" mean state is summarised by weather statistics that define "climate". Due to a strong desire to better understand climate variations and likely future climate change, there has been a rapid explosion in the use of descriptive statistical techniques in climate research. This talk will present a summary of some of the most commonly used approaches and will highlight some areas that could benefit from possible collaboration with statisticians.


Dr. Richard Chandler

Wednesday May 1, 2002

University College London

'Simple' Statistical Modelling of Climate Variability

Lecture Theatre D
School of Mathematics and Statistics,
University of St Andrews
The climate is a complex system. As a result of this the interpretation of climatological data is, on the face of it, a daunting task. This talk aims to demonstrate the enormous potential for progress when climatological expertise is combined with a knowledge of `standard' statistical methods (i.e. methods that would typically be encountered in an MSc level statistic course). The focus is primarily upon the use of Generalized Linear Models, along with standard time series modelling techniques. An attempt will be made to address some of the issues that arise when applying basic methods to such a complex system. The ideas will be illustrated with reference to a study of windspeed patterns throughout Northern Europe.

Thursday June 13th 2002, 5.45, Highland RSS AGM followed by:

6 pm

Stephen Senn

Thursday June 13th 2002

University College London

Individual Therapy: New Dawn or False Dawn?

Room 302,
Meston Building,
University of Aberdeen.
The sequencing of the human genome brings with it the hope that greater understanding of genetic components of disease will allow the more specific targeting of therapies. It has also been suggested that it will permit sponsors to run "cleaner" clinical trials with less variability and a consequent saving in patient numbers. However, we do not know how much of the variation in response that we see from patient to patient in clinical trials is genetic, because we rarely design the sort of trials that would allow us to identify patient-by-treatment interaction. Such interaction provides an upper bound for gene-by-treatment interaction for a group of patients studied since patients differ by more than their genes. On the other hand, however, the variability seen within a clinical trial may generally be expected to be less than the total variation that would be seen within a population. There is a related statistical issue to do with the interpretation of effects from clinical trials. This arises because there is confusion between experimental and sampling models of clinical research. It is concluded that we may have to pay careful attention to certain design features of clinical trials if we wish to make progress in this field.

Joint APERU/Highland RSS Meeting: Statistical Methods in Ecology Research


John Durban

Wednesday, 23rd October 2002

University of Aberdeen

Bayesian models and model selection for marine mammal population assessment

Zoology Lecture Theatre
Zoology Building
Aberdeen University
Effective management of marine mammal populations requires interaction with data from long-term monitoring programs. However, these data are generally sparse, requiring alternative techniques for inference about abundance and trends. A Bayesian approach will be presented for modelling photographic mark-recapture data. Issues to be addressed include estimating full posterior probability distributions for abundance estimates, and reducing uncertainty about abundance trends by the joint modelling of repeated estimates.




Steve Buckland, Ken Newman, Len Thomas and John Harwood

Wednesday, 23rd October 2002

University of St. Andrews

Formulating models for population dynamics

Zoology Lecture Theatre
Zoology Building
Aberdeen University
A general structure is proposed for formulating models for population dynamics. The structure allows complex models to be constructed using simple building blocks. These building blocks each represent a single process, such as survival, birth, assignment of gender or genotype to new-born animals, and movement. The rates corresponding to each process may be functions of covariates, allowing effects such as density dependence, predator-prey relationships, competition and habitat management to be incorporated. Inference proceeds using state-space models together with computer-intensive Bayesian methods. The methods are illustrated using a metapopulation of grey seals, for which a long, but incomplete, multivariate time-series of seal pup counts at the major UK colonies is available.

Commissioned analysis of surgical performance using routine data: The Bristol Inquiry
Nicky Best (Imperial College, London)
The public inquiry into paediatric cardiac surgery at the Bristol Royal Infirmary commissioned the speaker and colleagues to design and conduct analyses of routine data sources in order to compare surgical outcomes between centres. Such analyses are necessarily complex in this context, but were further hampered by the inherent inconsistencies and mediocre quality of the various data sources. Three levels of analysis of increasing sophistication were carried out. The reasonable consistency of the results arising from different data sources, together with a number of sensitivity analyses, led us to conclude that there had been excess mortality in Bristol in open-heart operations on children under 1 year of age. This talk will discuss some the statistical and data issues involved in carrying out this analysis, and highlight the potential role of statistics in future programmes for monitoring clinical performance.
Can we make Statistics count in Bioinformatics?
ERNST WIT (University of Glasgow)
6pm, Thursday 13 February 2003 Bioinformatics is a common denominator of many diverse activities, which range from database construction, via exploratory graphical analyses, to high-dimensional methods. What they share in common is (i)a focus on biological, often genetic, data and (ii) practitioners that often come from the numerical sciences, such as Computer Science and Engineering. Statisticians and Mathematicians have been slow to embrace this new field. For no good reason. In fact, Statistics is 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 high-dimensional 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.
Wednesday 7 May Lecture Theatre C, Mathematical Institute, North Haugh, St Andrews
3.00pm Dr. Robert Aykroyd (University of Leeds)
"A Bayesian approach to inverse problems in medicine, archaeology and industrial process engineering"
Many imaging problems can be defined as ill-posed or inverse problems: that is, either there is no unique solution or the solution does not depend continuously on the data. For inverse problems to be solved statistically, prior information must be included: solution based on the likelihood alone does not yield a stable solution. MCMC techniques can then be chosen to explore aspects of solutions. This approach will be described and illustrated using examples from medical SPECT imaging, archaeological magnetometry and industrial electrical impedance tomography. These are non-invasive techniques used to visualise physically inaccessible areas by taking measurements around the boundary. Reconstruction can allow improved medical diagnosis, the importance of potential archaeological sites to be assessed, and industrial processes to be efficiently monitored and controlled.
4.30pm Dr. Ruth King (University of Cambridge)
"Integrated Wildlife Populations: A Bayesian Approach"
Within recent years there have been increasing concerns over many different wildlife populations, resulting in the introduction of studies to observe these populations. Data may be collected in many different forms, including for example, capture-recapture data (of live animals), tag-recovery data (of dead animals), radio-tagging, location data and/or census data. Often, more than one form of data may be collected on a single population. With an increasing amount of data, new techniques need to be designed in order to incorporate all such information, thus removing incompatible conclusions that may result from considering the data sets independently, and potentially increasing the precision of the parameters. We shall consider in detail the analysis of joint ring-recovery and census data collected on the British northern lapwing. We shall consider a Bayesian analysis of the integrated data and discuss the increased information gained as a result of considering both forms of data jointly within a single analysis. Additionally, we shall show how we are able to consider a variety of "exact" models, removing the usual Normality approximations used to obtain classical estimates of the parameters.

Untangling sexual reproduction: two statistical problems in genetics
6pm, Thursday 5 June 2003 IAN WILSON (University of Aberdeen)
Meston Building, Room 302, Kings College, University of Aberdeen Sexual reproduction, an individual's genotype is a combination of the genes from both parents. A number of problems in statistical genetics can be thought of as disentangling the contributions. In this talk I shall outline two problems where inference about parental genotypes is important, one in ecology and another in the genetics of disease, and outline Bayesian methods for assigning genes to parents. In both cases I take a Bayesian approach using Markov chain Monte Carlo and discuss a method to deal with slow mixing.
9th September 2003

Peter Holmes

6pm, tea from 5.45pm

RSS Centre for Statistical Education

Assessment with a Purpose in Statistics

Room 302,
Meston Building,
University of Aberdeen.
Different forms of assessment can easily dominate teaching. In this talk I look at the fundamental purpose of all assessment, how it links to aims and learning outcomes, how assessment can both enhance and harm learning, and how the different levels of Bloom's Taxonomy can be matched with different forms of assessment in statistics.
9th October 2003

Malcolm Hall & Roger Humphry,

6 pm, tea from 5.45pm

Epidemiology Unit, Scottish Agricultural College, Inverness

Investigations into antibiotic resistance by the Veterinary Epidemiology Unit, Inverness

Room 302,
Meston Building,
University of Aberdeen.
This talk will describe some of the work undertaken by the Epidemiology Unit of the Scottish Agricultural College in Inverness. Sources and approaches to analysing data will be described with an emphasis on the antibiotic resistance of E. coli from cattle as our example. The results of a survey intended to estimate the prevalence of antibiotic resistance and its associations with farm management factors led to questions regarding how antibiotic resistance is defined and measured. We therefore devised, in collaboration with Graham Horgan (Biomathematics and Statistics Scotland), an alternative approach of defining antibiotic resistance based on non-linear regression and the cumulative density function of the normal distribution. Tests of this method for a limited number of samples are encouraging and it is intended to describe some of the issues and discuss possible refinements arising from these results.
Thursday 21st October 2004

John Norrie

5pm (tea/coffee from 4:45)

Director, Centre for Healthcare Randomised Trials (CHaRT), Health Services Research Unit, Aberdeen University

Randomised controlled trials (RCT) post EU Clinical Trials Directive - some issues for statisticians

Room WB27,
Health Services Research Unit,
Polwarth Building, Foresterhill,
The EU Clinical Trials Directive, along with other legislation and guidance, has given an exacting regulatory, ethical, and legal environment in which RCT are now conducted. This talk will give some observations on statisticians roles, from the perspective of an experienced triallist and statistician who has been involved in many RCT, both large and small, both publicy funded and ommercial, before and after the implementation of EU/CTD. The talk will include discussions of risk assessment of trials, some aspects of regulatory advice on both the design and analysis of RCT (via the Points to Consider series), and some thoughts on statisticians involvement in the monitoring of RCT.
Tuesday 16th November

Ottar N. Bjornstad


Penn State University, USA

The dynamics of spatially-extended populations: spatial synchrony and spatial correlation functions

Zoology Lecture Theatre, Zoology Building,
Aberdeen University (see
for directions)
Spatiotemporal dynamics can be understood by considering how local interactions affect patterns of spatial correlation (and cross-correlation), and how these in turn affect local dynamics. This feedback holds the key to understand (i) spatial synchrony, (ii) local dynamics and (iii) regional persistence. In particular, the local dynamics of consumer-resource systems are often unstable. Because of the inherent local instability, mobility introduce spatially transient associations. Persistence is achieved though spatial interactions. In this talk I focus on spatial correlation functions and how these relate to pattern and process. I subsequently outline theoretical predictions about correlation functions and cross-correlation functions. The correlation functions can be estimated from data. I discuss theory and data with respect to a range of case-studies.
Tuesday 16th November

Len Thomas, John Harwood, Stephen T. Buckland and Ken B. Newman

5.05pm (coffee at 4:50
following previous talk)

University of St Andrews

Use of sequential Monte-Carlo methods to fit and compare models for the dynamics of wild animal populations.

Zoology Lecture Theatre, Zoology Building,
Aberdeen University (see
for directions)
In a previous talk to this group, Steve Buckland showed how state-space models are a convenient and flexible framework for specifying stochastic models for the dynamics of wild animal populations and for the data available about these populations. We briefly review this work in the context of a spatially explicit model for the population of British grey seals, for which the available data include estimates of numbers of pups born each year in each breeding colony. Specifying these models is relatively easy, but fitting them is not. A number of techniques are available including Markov chain Monte-Carlo, Kalman filtering and sequential Monte-Carlo particle filtering (also called sequential importance sampling). We give an intuitive introduction to sequential Monte-Carlo methods and illustrate their application to the seal model. We also show how these methods can be used in a model selection problem, where the goal is to determine whether culls of seals around salmon farms could be causing the recent levelling-off in seal counts in some areas.
Thomas, L., S.T. Buckland, K.B. Newman & J. Harwood. In press. A unified framework for modelling wildlife population dynamics. Australian and New Zealand Journal of Statistics.
Newman, K.B., S.T. Buckland, S.T. Lindley, L. Thomas & C Fernández. In press. Hidden process models for animal population dynamics. Ecological Applications.
Thomas, L. & J. Harwood. 2004. Possible impacts on the British grey seal population of deliberate killing related to salmon farming. SCOS briefing paper 04/7.
Buckland, S.T., K.B. Newman, L. Thomas & N.B. Koesters. 2004. State-space models for the dynamics of wild animal populations. Ecological modelling 171: 157-175.
Thursday 10th March

Professor Peter Diggle

4.30pm (Tea from 4 pm)

Department of Mathematics and Statistics, Lancaster University

Spatio-temporal point processes: two problems, two approaches

Scottish Agricultural College, Sratherrick Rd,

Further details available from: Malcolm Hall, Scottish Agricultural College, Inverness (Tel: 01463 243030, email:
Spatio-temporal point process data arise naturally in a number of disciplines, including (human or veterinary) epidemiology where extensive data-sets are also becoming more common. In this talk I will describe two approaches to the analysis of spatio-temporal point process data, each motivated by a particular application as follows: 1) empiral modelling, exemplified by a log-Gaussian Cox process model for real-time monitoring of gastro-enteric disease in southern England; 2) mechanistic modelling, exemplified by a model due to Matt Keeling (Warwick University) for the 2001 UK foot-and-mouth epidemic. I will compare different approaches to inference, including maximum likelihood and a computationally simple partial likelihood alternative.
Thursday 21st April

David Braunholtz

4.30pm (Tea from 4.15 pm)


Causal models in Public Health research

Room 1.147 in the West block of the Polwarth
Building at the Medical School, Forester Hill,
Aberdeen (previously called WB20)
Map of Foresterhill site
Map of Medical School Building
The easiest Access is either slightly to the west of the letter C (once in door turn diagonally left and walk along glass sided corridor, then left again) or at the west-most end of that chain of buildings (once in door turn right, then left and room is only one on left).
Monday 13th June

Marian Scott

4:00 pm (Tea from 3:30 pm,
AGM: 5:10 pm)

University of Glasgow

Evidence-based environmental policy- what can we learn from routine environmental monitoring data?

Grampian Room, Macaulay Institute,
Craigiebuckler, Aberdeen
(please report to reception on arrival)
Many millions of pounds are spent on the routine collection of environmental data, which can then be used for a variety of purposes including: assessment of long-term changes either in natural conditions or resulting from anthropogenic activities, ascertaining the magnitude and impacts of accidental pollution and assessing compliance with environmental standards. The link to environmental policy is often less clear. In this talk, I review some of the statistical methods (spatial, temporal and spatio-temporal) used to analyse routine monitoring data using examples from air and water quality as illustrations and consider some examples of current environmental policy.

4pm,  Wed 5 October 2005

Grampian Room, Macaulay Institute, Craigiebuckler, Aberdeen (maps on )  

Adam Butler (Biomathematics & Statistics Scotland, Edinburgh) 

 Extreme value theory, climate change and coastal flood risk 

A number of scientific studies have suggested that climate change may be altering the frequency and magnitude of storms in the North Atlantic. Severe floods along the coastlines of the North Sea are typically associated with storm surges, so that any change in storm behaviour could be expected to have an impact upon coastal flood risk. We have used novel statistical models from the area of extreme value theory to analyse trends in storm surge levels over the past fifty years, and in this talk we give a broad overview of both the statistical methods and the oceanographic findings. We will assume no prior knowledge of extreme value theory, and will attempt throughout to demonstrate the wider applicability of extreme value methods within the environmental sciences.

tea at 3.30pm

4pm,  Wed 23 November 2005

Room 1:147, Polwarth Building, Foresterhill, Aberdeen
(see )  

Ian White (MRC Biostatistics, Cambridge) 

 Missing baseline data in randomised trials. 

I will briefly review the difficult problems of missing outcome data in randomised trials and missing covariates in observational studies. However, methods developed for these situations turn out not to be appropriate for dealing with missing baseline data in randomised trials. I will show that excluding participants with missing baselines is a bad option, but most other options, including simple mean imputation and a missing indicator method, are adequate.

tea at 3.30pm

(see )  

4.15. Joe N. Perry (Plant & Invertebrate Ecology Division Rothamsted Research, Harpenden)

Design of experiments and analysis of data concerning GM crops

My research on GM crops has largely involved collaboration on the design and analysis of the UK Farm Scale Evaluations (FSE) of genetically modified herbicide-tolerant (GMHT) crops. Results concerning the effects of herbicide management practices on farmland wildlife for three spring-sown crops (beet, spring oilseed rape and maize) were published in autumn 2003 and spring 2004; results from the fourth crop, winter oilseed rape, were published in Spring 2005. A very brief summary will be given of the results published to date. Also, two papers will be described that reassess the analysis, and the estimates made of statistical power.Work following on from the FSE, and other GM work will be described. Simple mathematical models can be used to show how the adverse effects of GMHT systems on the wildlife in sugar beet crops might be mitigated. Other studies involve mathematical modelling of relevance to the issue of coexistence between GM crops and organic or conventional crops. This focuses on issues around the distances proposed to separate GM and other crops and on spatial and temporal heterogeneity of crops in landscapes.

5.15. Tea / Coffee

5.30. Geoff Squire (Scottish Crop Research Institute, Dundee)

Genes and food webs in the GM crop trials

The results of the UK’s GM crop trials, which began in 2000, are here interpreted through the annual cycle of energy flow in arable fields. Growing herbicide tolerant crops altered the balance of solar energy absorbed by crops and by weeds to a small degree which nevertheless had important knock-on effects to invertebrate food webs and the persistence and movement of GM traits. Since the last GM crops were harvested in 2004, research teams have continued to measure the seeds, plants and genes at and around evaluation sites and to calculate the effects of GM herbicide-tolerant cropping on the arable ‘species pool’ across the UK. The findings, summarised in the talk, have lead to hypotheses on the dynamics of plants and genes, over time and over the landscape, that are now being tested using non-GM crop varieties. Finally, when compared with GM field experiments across Europe, the evaluations in the UK are shown to have several unique features that have greatly increased our understanding of the farmed habitat.

1 pm,  Wednesday, 19 April 2006

Med Chi Hall (Medico-Chirurgical Hall), Ground Floor, Polwarth Building, Foresterhill, Aberdeen [*]  

Jane Hutton (Warwick University)

Roles for statistics and statisticians in Ethics Committees

The need to consider whether the ethical legitimacy of a proposed medical experiment was recognised in the nineteenth century. The Nuremberg Code (1948) provides a clear framework within which to evalute ethics of human experiments. Most of the ten points require an understanding of risk and variation. This led to discussion of the relation between scientific quality and ethics, and some research ethics committees are now required to have a statistician as a member. More recently, the use of animals in experiments has been debated. This talk will summarise current practice for human medical research and developing practice for animal research.

[*] (see hospital page : hospital with medical school marked as A and the medical school page medical school page with Med Chi Hall marked as D. It is usually accessed via C or E as described below: If entering the Polwarth Building from the west, follow the main corridor east past the medical library, through one set of double doors and take the stairs to your left down one flight. The Med Chi Hall is through a single wooden door at the bottom of the stairs. The lecture room itself is to the left. If entering the Polwarth Building from the east through the main entrance, go through the foyer past the main stairs on the left and through an open set of double doors. Turn left signposted to Medical Microbiology. Follow this corridor to its end where there will be a single wooden door to the Med Chi Hall. The lecture room itself is to the left.)


SPEAKERS: Frank Critchley (Open University), Simon Wood (Bath)
DATE: WEDNESDAY 03/05/2006
TIME: 3pm
VENUE: Lecture Theatre D, Mathematical Institute, North Haugh, St Andrews
(for updates please go to:,
a map of the university can be found at:

3.00pm    Frank Critchley (Open University)
"Principal Axis Analysis " (abstract is given below)

4.00pm: Tea (in the Staff Room)

4.30: Simon Wood (University of Bath):
"Calanus in the North Atlantic: a simple approach to fitting a complex model."

The meeting will be followed at 6.30 p.m. by a 2-course meal at the Doll's House 
Restaurant in Church Square, St Andrews. Those wishing to come to the meal, please 
inform Peter Jupp, preferably by email ( before the end of Monday 
1 May (updates on the arrangement for the meal can also be found at:
Principal Axis Analysis
(Frank Critchley(1), Ana Pires(2) and Conceição Amado(2); Open University, UK(1) and IST, Lisbon(2))

Principal axis analysis rotates standardised principal components to
optimally detect subgroup structure, rotation being based on
preferred directions in the spherised data. As such, it is a
computationally efficient method of exploratory data analysis,
particularly well-suited to detecting mixtures of elliptically
contoured distributions. The ability of principal components itself
to perform as a cluster analysis method on some occasions, but not
others, is explained and illustrated. Links with a number of related
multivariate methods are explored. Examples are given throughout.
Further developments are briefly indicated. Overall, principal axis
analysis exemplifies the maxim: 'two decompositions are better than

Prof Adrian Bowman, University of Glasgow

"The use of additive models in environmental modelling".

Dr. Iain McKendrick, BioSS (Biomathematics & Statistics Scotland).

"E.coli 0157: statistics and mathematical modelling to integrate epidemiology and clinical trials".

Time: 3pm - 5.15pm, Wednesday 7th June 2006.
Venue: Epidemiology Research Unit, SAC, Drummondhill, Inverness, IV2 4JZ.
Map: Here
Queries: contact

Jonathan Stern (University of Bristol)
How can we motivate medical students to learn about medical statistics?

DATE: Monday, 9 October 2006
TIME: 12:30
VENUE: Med Chi Hall (Medico-Chirurgical Hall), Ground Floor, Polwarth Building
(see hospital page with medical school marked as A and the medical school page with Med Chi Hall marked as D. It is usually accessed via C or E as described below:
If entering the Polwarth Building from the west, follow the main corridor east past the medical library, through one set of double doors and take the stairs to your left down one flight. The Med Chi Hall is through a single wooden door at the bottom of the stairs. The lecture room itself is to the left.
If entering the Polwarth Building from the east through the main entrance, go through the foyer past the main stairs on the left and through an open set of double doors. Turn left signposted to Medical Microbiology. Follow this corridor to its end where there will be a single wooden door to the Med Chi Hall. The lecture room itself is to the left.)
For many medical students, medical statistics appears to be a subject that is peripheral 
to the core knowledge that is needed to practise as a junior doctor. On the other hand, 
it is widely acknowledged that medical practice should be evidence-based, and reading the
medical literature requires an ability to interpret the results of statistical analyses 
as well as more general critical appraisal skills. I will discuss strategies to persuade
students that, while they may never love medical statistics, knowledge of the subject is 

David Kerridge (emeritus of Aberdeen University)
Paul Garthwaite(Open University)
Peter Fayers (Department of Public Health, Aberdeen University)


DATE: Tuesday, 5 December 2006

TIME: 12:30

12:30-13:30: Sandwich lunch
13:30-14:00: Welcome + Introduction
14:00-14:45: David Kerridge
14:45:15:30: Paul Garthwaite
15:30-16:00: tea break
16:00-16:45: Peter Fayers

René Holst (DIFRES,Denmark)
Analysis of clustered binary data using estimating equations with an application to gear selectivity

DATE: POSTPONED (was 8 Feb 2007)
TIME: 4pm (tea/coffee available from 3:45pm).
VENUE: FRS Marine Laboratory, Aberdeen (for directions, see:
The selective properties of trawl fishing gear are commonly assessed by mounting a small mesh cover over the trawl codend. Fish entering the gear will therefore end up in the codend or in the cover. An experiment typically runs over several hauls, forming a clustered structure on the data. Data from such experiments are thus binary clustered data. Traditional approaches to the analysis of mixed effects models involve high dimensional integrals for handling the unobserved random effects. Except for the normal case these are intractable and are typically addressed by various approximations. In the talk I’ll briefly review some of the more important methods and their philosophical ideas. Secondly I’ll present a method based estimating equations using BLUP to predict the random effects. Asymptotic properties and computationally aspects will be discussed. The method will be demonstrated by an application to some gear selectivity data.

Sue Welham
Rothamsted Research
Evaluation of models for late-stage variety evaluation trials

DATE: 20 March 2007
TIME: 4pm (tea/coffee available from 3:40pm).
VENUE: Rowett Research Institute, Aberdeen (see directions)
The primary aim of crop variety evaluation programs is to reliably predict the future performance of potential new varieties relative to existing commercial varieties. This is achieved through multi-environment trials (METs), that is, series of field trials conducted across a range of geographic trial locations and possibly over several years. Varieties are tested for a range of agronomic traits with the aim of recommending superior varieties for commercial release. Usually, a program of METs is established to take varieties through from the initial selection of potential breeding lines to eventual commercial release and recommendation to farmers (late-stage variety evaluation trials). In this talk we are concerned with METs in the latter stages of this process.

Many models have been proposed for MET data, which can be broadly classified as one-stage or two-stage analyses (Smith, Cullis & Thompson, 2005). In a one-stage analysis individual plot data from all trials is combined in a single analysis. In a two-stage analysis variety means are first obtained from the separate analysis of individual trials (stage I), and are then combined in an overall mixed model analysis (stage II). The stage II analysis may be unweighted or weighted to reflect the relative precision of variety means from each trial. In each case, a linear mixed model may be constructed to describe the structure of the data, and this model is usually fitted using REML estimation. Within this context, variety by environment effects are commonly modelled using either simple variance components or a factor analytic model. Within-trial variation may be modelled using either a spatial model or blocking factors corresponding to the design.

This talk uses the characteristics of real MET data-sets from Australia and the UK as a basis for a simulation study to evaluate these different mixed modelling approaches. Results and their implications for the analysis of late-stage variety testing programs are presented.

Smith, AB, Cullis, BR & Thompson R (2005) The analysis of crop cultivar breeding and evaluation trials: an overview of current mixed model approaches. J. Agric. Sci., 143, 1-14.

3.00pm    Chris Jones (Open University)
"Don't Mention Stonehenge!  The Statistician's Side of the Story of a
Ten Year  Collaboration between Archaeology, Earth Sciences and
Statistics" (abstract is given below)

4.00pm: Tea (in the Staff Room)

4.30:  John Hinde (National University of Ireland,Galway):
"Random Effects, Mixtures and NPMLE"

The meeting will be followed by a meal (to be arranged).


             Wednesday 2 May

             Lecture Theatre D

3.00 p.m.     Professor Chris Jones (Open University)

         Don't Mention Stonehenge!
     The Statistician's Side of the Story of a Ten Year
Collaboration between Archaeology, Earth Sciences and Statistics


The subtitle is the short version of the abstract! In 1994,
archaeologist Olwen Williams-Thorpe helped me out by appearing on an
Open University television programme associated with a Statistics
course. In return, I agreed to help her out with a bit of statistical
consultancy. More than 10 years - and 10 papers - later, this
collaboration has come to an end. In this talk, I'll review this
successful collaboration between archaeology, earth sciences and
statistics. Our work focussed on the magnetic and chemical properties of
stone artefacts and what these quantitative measurements could tell us
about their provenance: from Roman columns quarried in the Egyptian
desert to British Bronze Age stone axes sourced from the Whin Sill in
Northern England and the Preseli Mountains in South Wales. The talk will
be broad-brush, yet at times selective with an emphasis on the
statistics, anecdotal, and perhaps even occasionally amusing, rather
than going into the scientific outcomes in depth (particularly as I am
no archaeologist or geologist).


4.30 p.m.   Professor John Hinde (National University of Ireland,

         Random Effects, Mixtures and NPMLE


The use of mixed models with normally distributed response and random
effects is well worked out and available in many software packages. For
non-normal response data the situation is less clear and there are many
different approaches being studied. Within the framework of generalized
linear models, one natural way of proceeding is to include additive
random effects in the linear predictor. These random effects can be used
to account for an additional level of individual variability
(overdispersion); a shared random effect for the additional variance
component in two-stage sample designs; longitudinal dependence in
repeated measures designs; spatial dependence in disease mapping; etc.
A common assumption for the random effects is that they are normally
distributed, giving the generalized linear mixed model, where maximum
likelihood estimates can be obtained using the EM algorithm and
essentially fitting a mixture model. By a simple extension of this
approach, it is possible to relax the assumption of normality and obtain
a nonparametric maximum likelihood estimate for the random effects
In this talk we will give a brief introduction to the methodology and
discuss several illustrative examples using the recent implemented R
package npmlreg.

René Holst (DIFRES,Denmark)
Analysis of clustered binary data using estimating equations with an application to gear selectivity

DATE: Thursday, May 2006

TIME: 3pm

VENUE: FRS Marine Laboratory, Aberdeen (for directions, see:

The selective properties of trawl fishing gear are commonly assessed by mounting a small mesh cover over the trawl codend. Fish entering the gear will therefore end up in the codend or in the cover. An experiment typically runs over several hauls, forming a clustered structure on the data. Data from such experiments are thus binary clustered data. Traditional approaches to the analysis of mixed effects models involve high dimensional integrals for handling the unobserved random effects. Except for the normal case these are intractable and are typically addressed by various approximations. In the talk I’ll briefly review some of the more important methods and their philosophical ideas. Secondly I’ll present a method based estimating equations using BLUP to predict the random effects. Asymptotic properties and computationally aspects will be discussed. The method will be demonstrated by an application to some gear selectivity data.

Natalia Bochkina (Edinburgh)
Objective and subjective approaches to analysis of differential expression in microarray data (joint work with Alex Lewin and Sylvia Richardson)

DATE: Tuesday, 30 October 2007

TIME: 4pm (Tea from 3:45)

VENUE: Strathcona Lecture Theatre, Rowett Research Institute, Greenburn Road, AB21 9SB (See details at: The Strathcona lecture theatre can be found within Strathcona Hall. This is the building to your right (opposite from the Reid Library/Reception), when you come to Greenburn Road from the A96.)

We consider a problem of comparing the means in two conditions (such as disease versus healthy samples) for a large number of variables simultaneously given a small number of replicates, with motivation coming from analysis of gene expression data. Bayesian modelling of such data allows to "borrow strength" across variables to stabilise variance estimates given a small number of replicates, by using an exchangeable hierarchical prior for variable-specific variances.
With regard to the mean, with consider two types of priors: noninformative ("objective") and mixture priors. To define a mixture prior, we use prior knowledge that there are 3 groups of variables: with equal means (the null hypothesis), with the mean in condition 1 greater than the mean in condition 2 (overexpressed) and with the reverse (underexpressed). Thus, we use a three component mixture distribution to model the difference between the means, with components being an atom at zero and gamma distributions with support on positive and on negative semiline. We study sensitivity to the choice of prior on simulated data and compare models with different priors using posterior predictive checks.
Since the mixture model can be sensitive to the choice of prior, we also consider a Bayesian model with noninformative prior for the means. To find variables with unequal means, we propose adaptive interval hypothesis testing where the interval depends on variability of each variable. Since we are testing a large number of hypotheses simultaneously, we also propose an estimator of the false discovery rate which allows to control the number of false positives. The adaptive interval hypothesis testing can be extended to compound hypotheses, i.e. those involving more than one parameter.
These approaches will be illustrated on gene expression data sets produced by BAIR consortium (

Natalia Bochkina (University of Edinburgh)

Objective and subjective approaches to analysis of differential expression in microarray data (joint work with Alex Lewin and Sylvia Richardson)

DATE: Tuesday, 30 October 2007

TIME: 4pm (Tea from 3:45)

VENUE: Strathcona Lecture Theatre, Rowett Research Institute, Greenburn Road, AB21 9SB (See details at: The Strathcona lecture theatre can be found within Strathcona Hall. This is the building to your right (opposite from the Reid Library/Reception), when you come to Greenburn Road from the A96.)

We consider a problem of comparing the means in two conditions (such as disease versus healthy samples) for a large number of variables simultaneously given a small number of replicates, with motivation coming from analysis of gene expression data. Bayesian modelling of such data allows to "borrow strength" across variables to stabilise variance estimates given a small number of replicates, by using an exchangeable hierarchical prior for variable-specific variances.
With regard to the mean, with consider two types of priors: noninformative ("objective") and mixture priors. To define a mixture prior, we use prior knowledge that there are 3 groups of variables: with equal means (the null hypothesis), with the mean in condition 1 greater than the mean in condition 2 (overexpressed) and with the reverse (underexpressed). Thus, we use a three component mixture distribution to model the difference between the means, with components being an atom at zero and gamma distributions with support on positive and on negative semiline. We study sensitivity to the choice of prior on simulated data and compare models with different priors using posterior predictive checks.
Since the mixture model can be sensitive to the choice of prior, we also consider a Bayesian model with noninformative prior for the means. To find variables with unequal means, we propose adaptive interval hypothesis testing where the interval depends on variability of each variable. Since we are testing a large number of hypotheses simultaneously, we also propose an estimator of the false discovery rate which allows to control the number of false positives. The adaptive interval hypothesis testing can be extended to compound hypotheses, i.e. those involving more than one parameter.
These approaches will be illustrated on gene expression data sets produced by BAIR consortium (

Peter Hall (Melbourne)

Robustness of multiple hypothesis testing procedures against dependence

DATE: Wednesday, 5 December 2007

TIME: 4pm

VENUE: Room 115, Institute of Medical Sciences, ForesterHill Campus, Aberdeen University

Problems involving classification of high-dimensional data, and `highly multiple' hypothesis testing, arise frequently in the analysis of genetic data and complex signals. Their theoretical elucidation raises challenges, however. We address this issue by interpreting small samples of high-dimensional data as small numbers of replicates of long segments of nonstationary time-series. Depending to some extent on how erratic the time-series are, important features of classifiers, or of multiple hypothesis testing procedures, can be accessed by exploring properties of time-series models. For example it can be shown that, in the context of multiple hypothesis testing, the assumption of independence is much less of an issue in high-dimensional settings than in conventional, low-dimensional ones. This is particularly true when the null distributions of test statistics are relatively light-tailed, for instance when they can plausibly be based on Normal approximations. Similar arguments can be employed to explore other aspects of the analysis of high-dimensional data.

Dr Ben Cooper (Statistics, Modelling and Bioinformatics Department Centre for Infections, Health Protection Agency, London)

The analysis of hospital infection data using a mechanistic model

DATE: Friday, 8 February 2008

TIME: 4pm (tea and coffee from 3:45)

VENUE: Room 1.029, Polwarth Building, Medical School, Foresterhill, Aberdeen University (This is the main Medical School building accessed from Ashgrove Road West or on foot from Cornhill Road. A map is available at this website . Clicking on the letter A will zoom in on the separate buildings. On the zoomed map, the talk will take place on the first floor of building E.)

Joint Aberdeen University/RSS Highland Local Group Meeting:
Statistical Methods in Population Ecology

Stijn Bierman (Biomathematics & Statistics Scotland, Edinburgh)
The analysis of data from two contrasting types of large-scale ecological surveys

Thomas Cornulier (Aberdeen Univ/BIOresc, Mammal Research Inst, Polish Academy of Sciences)
New approaches for predicting farmland bird population dynamics with data and process uncertainty

DATE: Wednesday 30 April 2008
2:15 Stijn Bierman
3:15 Tea/Coffee
3:30 Thomas Cornulier
VENUE: Zoology Lecture Theatre, Zoology Building, Aberdeen University (see here for directions)

Stijn Bierman, Chris Glasbey, Adam Butler, Glenn Marion, Ingolf Kuhn
(Biomathematics & Statistics Scotland, Edinburgh):
The analysis of data from two contrasting types of large-scale
ecological surveys

We present the results of research into suitable statistical approaches
for the analysis of data arising from two contrasting types of
large-scale ecological surveys, each with different questions of
interest: biological atlas data and data on population sizes from a
sample of spatial locations. In the case of biological atlases, data
collected by large numbers of volunteers are aggregated over both space
and time in an attempt to give a coarse but as complete as possible
representation of the distribution of species over large geographical
areas. The statistical interpretation of these data is complicated
because species are not perfectly detected. We present a number of
model-based approaches for the analysis of these data, and illustrate
these using the German Atlas of Vascular plants. In contrast, many
large-scale ecological survey schemes collect data on population sizes
from a sample of spatial locations within a region, and use this sample
to estimate (trends in) regional population densities. These data can
typically be recorded with little error but only a small sample of the
total area of interest will be available. Here, we use a variety of data
sets with complete enumerations of spatial populations to explore the
relative efficiency of combinations of spatial survey designs and
design-based estimators of standard errors of sample means.

Thomas Cornulier (Aberdeen University/BIOresc, Mammal Research
Institute, Polish Academy of Sciences):
New approaches for predicting farmland bird
population dynamics with data and process uncertainty
Mechanistic models should predict population dynamics better than
descriptive ones under new environments. However, there is often
considerable uncertainty in the parameters estimated from field data and
in the demographic processes for a particular species. Within this
UKPopNet project, we aimed at aimed at harnessing new estimation methods
to fit mechanistic models to a range a data types. Using Bayesian
approaches, we improved existing models for estimating demographics
parameters such as survival. We also developed a new approach based on
mixture models for estimating so far unavailable parameters such as the
number of breeding attempts. Using state-space models, we integrated
these estimates and manipulated model structure in order to investigate
different regulation or limitation processes, e.g. density-dependence
mechanisms, variation in the number of breeding attempts, habitat loss.
We will show how data and process uncertainties affect predictions for
the yellowhammer, a declining farmland bird in the UK. 

John Aston (University of Warwick)
The statistical analysis of PET time courses

DATE: Monday 19 January 2009
Time: 4pm
VENUE: Room 115 on the first floor of the Health Sciences Building, Forester Hill campus of Aberdeen University (The Health Sciences building is beside A on this map and is building H on this map.
There is a receptionist on the front desk of that building. Parking immediately beside this building will not be possible. Parking in the car park off Ashgrove Road West (beside the building marked A) should be possible at 4pm and the restrictions for on-street parking stop at 4pm.)

Positron Emission Tomography (PET) is a powerful technique for imaging the human 
brain. One of its major strengths is that it can examine the concentration of 
chemicals in brain regions over time, chemicals associated with diseases such 
as Alzheimer's and Parkinson's. However, in these situations, the measured PET 
time course is non-linear in nature and techniques are needed to deal with these
non-linearities. In addition, while PET time courses tend to have a small number 
(around 30) measurements in time, there are often several million measurements 
in space which comprise the image volume. Therefore, it would seem advantageous 
to use the spatial information as well as the temporal information in the analysis.

This talk will examine various methods of PET time course analysis, starting at 
the traditional point-by-point compartmental model, moving through techniques 
for model selection, and finally examining recent results in the use of 
non-parametric techniques for time course smoothing. It will be shown that by 
considering both space and time, more effective analysis can be achieved.

Darren Green ( Institute of Aquaculture , University of Stirling)
Structured Populations, Networks and Epidemic Modelling

DATE: Thur 5 March 2009
Time: 4pm
VENUE: Fisheries Research Services, Victoria Road, Aberdeen


Wednesday 13 May

Lecture Theatre D,
Mathematical Institute,
North Haugh, St Andrews
(for updates please go to , and a map of the university can be found at )

3:00: Paul D. Baxter (University of Leeds)
"Adventures in Applied Statistics"
"The best thing about being a statistician is that you get to play in 
everyone's backyard", so said John Tukey. The most enjoyable aspect of my 
career so far has been the opportunity to work with different disciplines, 
using real world problems to motivate new methodology and transfer existing 
ideas to new contexts. In this talk I will discuss work in three areas I 
have been involved with: criminology, ecology and environmetrics.
4:00: Tea (in the Staff Room)
4:30: Dr Duncan Lee (University of Glasgow)
" How does air pollution affect human health?"
The relationship between air pollution exposure and human health has been 
extensively investigated during the last twenty years, focusing on both 
the acute and chronic health effects. Acute effects describe the impact 
that a few days of high exposure have on human health over the following 
week, and are estimated by population level studies. These studies regress 
daily counts of mortality or morbidity events from a given population, 
such as those living in a city, against air pollution concentrations and 
other risk factors, including measures of meteorology and longer-term trends. 
In contrast, chronic effects are the consequences of long-term exposure 
over numerous years, and can be estimated by cohort and population level 
studies. The latter regress yearly counts of mortality or morbidity events, 
from contiguous small areas, against average air pollution concentrations 
over the preceding few years, allowing for the effects of demography and 
socio-economic deprivation. This talk gives an overview of both acute and 
chronic air pollution and health studies, including the standard modeling 
approaches and the methodological extensions that have been proposed. 
A number of examples of these studies are presented, focusing specifically 
on data from Great Britain.

Marianne Defernez

Institute of Food Research, Norwich Research Park, Colney Lane, Norwich NR4 7UA, UK

Analysing metabolomics' data: A challenging task

DATE: Tues 23 Feb at 3pm
VENUE: Room 3.052 of the Polwarth Building (*)

Metabolomics is truly multi-disciplinary: It requires a multitude of skills for the research to be successful, ranging from spectroscopic and statistical expertise to in-depth knowledge of the biological systems under study. In this talk I will focus on the data analysis and show why and how some of the statistical methods used have evolved since the term 'metabolomics' was first coined twelve years ago. This will provide an opportunity to highlight current trends and limitations. It will also demonstrate that the data analysis should not be seen simply as the concluding part of an investigation; it is in fact greatly influenced by upstream steps. In particular the data pre-treatment necessarily to make the raw data suitable for the statistical analysis itself has a huge impact on the outcome, and I will discuss this issue for both NMR, and LC and GC-MS data. The talk will show the benefit of having a better appreciation of the overall procedure required for investigations in this field.

(*) This room is on the third floor and at the Eastern end of the Polwarth Building above main entrance door. A lift is available. No swipe cards will be needed to access this building in the afternoon. A map is available, and the "hybrid" option does show the shape and colour of the buildings for those unfamiliar with the site.

RSS Highlands Local Group Meeting: Choice Experiments in Health and Environment


Thursday 15th April, 2pm


Macaulay Suite B, The Macaulay Land Use Research Institute, Craigiebuckler, Aberdeen.



Speaker: Professor Mandy Ryan, Health Economics Research Unit, University of Aberdeen


Title: Valuing benefits in the provision of health care: using discrete choice experiments to go beyond clinical outcomes


Abstract: Given the absence of a market for health care, a key challenge in health economics is the valuation of benefits from alternative health care interventions. The quality adjusted life year (QALY) has become the main valuation technique at the policy level. For example, it is the valuation method recommended by bodies such as the National Institute for Health and Clinical Excellence (NICE) and the Scottish Medicines Consortium (SMC). QALYs are also commonly employed in randomised controlled trials to assess the value of alternative health technologies. The QALY focuses on health outcomes, ignoring what may be termed ‘patient experience factors’. Where patient experience factors are important reliance on the QALY may lead to inefficient decisions.  This is demonstrated with a study looking at extending the role of the pharmacist in the delivery of medicines. Whilst this study demonstrated no significant difference in QALYs across the two arms of the trial, there was a difference in satisfaction levels. A follow-up discrete choice experiment (DCE) was used to value such differences. The DCE methodology is introduced and discussed, with its application to going beyond QALYS demonstrated. It is argued that, for efficient decisions regarding the allocation of scarce health care resources to be made, benefits beyond QALYs must be valued, and the DCE approach is a potentially useful methodology.




Speaker: Klaus Glenk, Scottish Agricultural College


Title: Incorporating and modelling delivery uncertainty in benefit assessments based on stated preference methods


Abstract: The economic valuation of benefits resulting from environmental policies and interventions typically assumes that environmental outcomes are certain, whilst they are often uncertain. We propose a methodological approach to incorporating delivery uncertainty into benefit estimation based on stated preference methods. Delivery uncertainty was explicitly included into the study design of a choice experiment survey on land-based climate change mitigation as the risk that a proposed mitigation project may fail to deliver emission savings. Issues of modelling delivery uncertainty as separate additive effects of risk and outcome, or using a multiplicative weighting function (linear or following prospect theory) will be discussed. Model results suggest that delivery uncertainty can have a large impact on stated preference estimation of benefits of public programmes, and hence conclusions drawn from ex-ante environmental cost-benefit analyses that make use of such benefit estimates.


Details of the joint RSS Highland Local Group meeting with St. Andrews University which was held on Wednesday 5th May 2010 in Lecture Theatre C of the Mathematical Institute, North Haugh, St. Andrews from 3pm.

The meeting will include presentations by Professor Paul Blackwell (University of Sheffield) and Professor Håvard Rue (Norwegian University of Science and Technology). Paul will discuss how MCMC can be used to infer key features of animal movements using long term GPS data. Following after tea and biscuits, Håvard will describe how some Bayesian models can be analysed using rapid integrated nested Laplace approximations rather than MCMC. Abstracts for both talks are given below.

The meeting will be followed at 6.30 p.m. by a meal to which all are welcome. Individuals wishing to attend the meal are asked to let the local organiser Janine Illian know, preferably by Friday April 30th, so that an appropriate table can be organised.

It promises to be an interesting afternoon, with presentations and discussion from two well-known statisticians. A map of the university can be found here .


Paul Blackwell

Department of Probability and Statistics, University of Sheffield
Bayesian Inference for Animal Movement Data

Radio-tracking, the use of GPS collars etc, to track individual animals is an increasingly important approach to monitoring, understanding and managing wildlife populations. These methods form an vital source of data on movement, behaviour and habitat use, but the complex structure of the data means that sophisticated statistical models of the underlying movement process are required to enable meaningful analysis. I will describe a rich class of continuous-time models which aims to capture key features of realistic patterns of animal movements. Complex behaviours can be driven not only by internal mechanisms but also by the heterogeneous environment of the animal, so the models can allow different movement processes in regions of the habitat, with boundaries which may be observed or may be unobserved and may therefore need estimating. The approach to inference is Bayesian using Markov chain Monte Carlo techniques, allowing us to estimate parameters of the process and the environment, and to predict or interpolate movement. I will illustrate the models using long-term GPS data on wild boar, courtesy of the wildlife ecology and management group within the Food and Environment Research Agency, which display clear spatial and temporal heterogeneity.

Håvard Rue

Department of Mathematical Sciences, Norwegian University of Science and Technology
Bayesian computing with INLA

Many models in statistics can now to analysed using quick-to-compute integrated nested Laplace approximations (INLA) instead tedious MCMC sampling. In this talk I will present the main ideas of this approach, which models it can deal with and demonstrate how the analysis can be done in practice from within R. The software is available from .
Dr. Helen Brown (University of Edinburgh) is to give a seminar entitled 'Multilevel (Mixed) Models and Applications in Medicine and the Social Sciences' on Monday 29th November at the Institute of Medical Sciences in room 1.147 of the Polwarth Building, Forresterhill between 1pm and 2pm. Helen Brown, the senior statistician for the the 'Applied Quantitative Methods Network (AQMeN)', has a long-experience of mixed models and is an author of the book Applied Mixed Models in Medicine, one of the key texts in the field.

It is very likely that we will be joined by members of the AQMeN network in Aberdeen. AQMeN is a three year project funded jointly by the Economic and Social Research Council (ESRC) and the Scottish Funding Council (SFC) as part of the ESRC Quantitative Methods Initiative whose main objective is to build capacity in the use of intermediate and advanced level quantitative methods amongst the social science community in Scotland and beyond.

Helen has provided the following abstract. 'Multilevel or mixed models are becoming an increasingly popular method of data analysis. There are many situations where improved analyses can be obtained compared to more conventional approaches. This presentation will introduce multilevel models, discuss situations where they are appropriate, and show several applications in medical and social science settings'.

It would be helpful if you could send a brief email to Gordon Prescott (, Chair of the RSS HLG, who has organised the seminar if you hope to come. But don't let not sending an email prevent you from making a last minute decision to attend. All are, as ever, welcome.

The room is at the Western end of the Polwarth Building and access doors are available from the North beside the Medical Library or from the South up a flight of stairs opposite the revolving door and rotunda of the ARI. Please contact Gordon Prescott ( for details if wheelchair access is required.

There will be an RSS meeting in Inverness on Wednesday 1st December. It is being held at The Green House, Beachwood Business Park at 3:30pm in the Lecture Theatre. Professor Denis Mollison (Heriot-Watt University) is giving a talk titled 'Wave Energy: Making Best Use of a Variable Resource'. It would be helpful (but not at all compulsory) to let the local organiser, Roger Humphry, know if you are planning to attend ( so that he can plan accordingly. Tea and coffee is available from 2:45, and as is usual with our Inverness meetings we will head for a cost-efficient place after the talk not too far from the railway station to eat and chat for those wishing to do so. All are welcome, particularly statisticians; although the seminar is attracting a lot of attention from the local academic and business community there are relatively few professional statisticians working in the area.

Abstract: Scotland has one of the best wave energy resources in the world, and has been a leader in research on its exploitation since the 1970s: research that is at last leading to sea-going devices that could exploit a significant proportion of the resource. The talk will discuss some of the wide range of associated statistical problems, such as:
- estimation of the resource, both large and small scale, and its variability on various scales
- design of wave power devices to optimise productivity
- selection of representative sets of spectra, and their use in tank tests and productivity estimates
- integrating wave energy together with other renewables into a reliable electricity supply

Peter Challenor of the National Oceanography Centre (a joint venture between the Natural Environmental Research Council and The University of Southampton) agreed to lead our first Royal Statistical Society Highland Local Group meeting of the season.

The subject of his talk, which has broad applicability and includes specific examples from both climate and oceanography, is entitled "The Statistical Analysis of Complex Numerical Models in the Environmental Sciences". An abstract for the talk is given at the end of this notice.

The meeting will be held on Tuesday the 25th October at the The James Hutton Institute, Aberdeen. The talk will start at 3:30 pm, although coffee (and biscuits!) and a valuable chance to chat will be available from 3:00 pm. All are welcome to attend from 3:00 pm onwards, whether RSS members or not, so please feel free to pass the word around your colleagues.

The meeting will conclude with a (hopefully) brief AGM for RSS Highland Local Group members and will be the subject a later notice. Individuals not attending the AGM are welcome to continue their discussions while this is taking place. The abstract for Peter's talk is given below:


The Statistical Analysis of Complex Numerical Models in the Environmental Sciences Complex, usually deterministic, numerical models are now widely used in environmental science. Predictions from such models are normally presented without any measure of uncertainty. In this talk I shall show how statistical methods can be used to not only estimate the uncertainty on model predictions but also to carry out sensitivity analysis and to calibrate the models. Our basic tool is the emulator. The emulator is a statistical approximation to the full numerical simulator; we use a Gaussian process. We build the emulator from an initial designed set of model runs. Because it is fast to run and we know its statistical properties, the emulator we can be used for inference, either analytically or using Monte Carlo methods. Drawing on examples in climate and oceanography I will illustrate the construction and use of statistical emulators from the design of experiments to the estimation of extreme values.
It is a pleasure to announce that Professor Dave Collet (Director of Statistics and Audit at NHS Blood and Transplant) is visiting Aberdeen to lead our next RSS Highland Local Group meeting. Dave Collet is the Director of Statistics and Audit at NHS Blood and Transplant and is the author of well known statistical texts on modelling binary data and survival data.

The meeting will comprise two parts. The first part is a talk titled 'Statistical Issues on the Allocation of Livers for Transplantation' which will describe the use of statistical survival modelling to ensure the fair and unbiased allocation of organs. Such modelling takes into account informative censoring and is likely to be of interest to many statisticians. The abstract for this talk is given at the end of this email.

Dave has then agreed to a second part in which he will respond to any questions regarding survival analysis that we may have.

The meeting will be held on the Tuesday 29th November with the talk starting at 1pm in room 115 of the Health Sciences Building, Institute of Medical Sciences, at Forresterhill, Aberdeen. Those driving to the talk are advised to contact the meeting organiser, David McLernon (, for one of the limited number of visitor parking spaces at least a week prior to the talk. But don't let that put you off coming, please.

All are welcome to attend, whether members of the RSS or not, so feel free to spread the word around anyone who may be interested.

Liver transplantation has become a well-established treatment for liver failure, but the shortage of suitable organs from deceased donors limits the number of transplants in the UK to less than 700 a year. There is therefore a need for procedures that ensure that donated organs are allocated in a fair, transparent and unbiased manner. New liver allocation schemes are being developed based on statistical models for survival with and without a transplant. The development of these models needs to take account of informative censoring, and so methods for exploring the extent of informative censoring, and allowing for it in the modelling process, will be described. The number of livers available for transplantation can also be increased by splitting a whole liver to provide for two transplants. An analysis to compare outcomes following split liver transplantation with remaining on the waiting list for a whole liver will also be presented.
Our next RSS Highland Local Group gathering will be a discussion meeting on ‘Statistics and Policy’. The meeting will be held on Wednesday 8th February at the Marine Laboratory, Marine Scotland Science, 375 Victoria Rd, Torry, Aberdeen. Tea and coffee (with biscuits) is available from 3pm, with the formal meeting starting at 3:30 and ending around 5:00 pm.

Each of three speakers, with practical experience of providing statistical advice to policy makers from different disciplines, will present a short case-study describing some of their experience. The speakers come from different parts of our local area, and include Mr. Mike Lonergan (Sea Mammal Research Unit, University of St. Andrews), Mr. Cameron Thomas (Planning and Development Service, The Highland Council), and Professor Craig Ramsay (Health Services Research Unit, University of Aberdeen). Following the talks we will, together, explore what constitutes good practice and share techniques to overcome some of the practical problems that we face. There is considerable experience in this subject within the Highland Local Group area, and this meeting is intended to bring together some of this expertise. I hope to prepare a one-page summary document from the meeting which will be sent to all those attending.

It is very likely that we will conclude the meeting with a trip to a local pub and a bite to eat for anyone wishing to participate.

All are welcome to attend this meeting whether RSS members or not. So please feel free to spread the word around.
It is a pleasure to announce that our next RSS Highland Local Group meeting will comprise a talk by Professor Ian Diamond (University of Aberdeen) entitled The development of the use of statistics in the census and the future of the census.

The talk will provide a brief history of the census, looks at the development of the acceptance that census counts were subject to error and describes strategies used to make census estimates. Professor Diamond has considerable experience of working with the census, in particular, working on the problem of undercounting and inferring associated characteristics.

The meeting will be held on Thursday 10th May at 3.30-5pm and will be held in the conference room, the Suttie Centre, University of Aberdeen, Foresterhill.

All are welcome attend, including non-members, so feel free to pass the word around your collegues and friends. It would be helpful, but not necessary, if you could contact Gordon Prescott ( if you wish to attend.
It is a pleasure to announce that the RSS Highland Local Group will be hosting Dr Jonathan Cook (University of Aberdeen) who will be presenting a review of methods for specifying the target difference in randomised controlled trials. Dr. Cook is responsible for methodological research at the Health Services Research Unit, and the subject of his talk clearly has an applied and broad application. The work he is describing is a collaborative project with the Universities of Aberdeen, East Anglia, Glasgow, Newcastle and Oxford (in the UK), and the National University of Ireland in Galway and the Ottawa Research Institute in Canada (elsewhere). A title and abstract for his talk is appended at the bottom of this announcement. All, including non-members, are very welcome to attend this event.

After the discussion following Jonathan's talk we will hold our (hopefully) brief local group Annual General Meeting to which all HLG members are very welcome.

The meeting will be held on Wednesday 31st October at 3:30 pm (with refreshments available from 3:00 pm onwards) at the Med-Chi Hall which is within the Polwarth Building of the Medical School at Forresterhill, Aberdeen . A map giving the location of the venue can be accessed via the link here .

It would be helpful, but is not essential, if you would let Dr. Gordon Prescott ( know if you are planning to attend to gauge the amount of refreshments needed. But don't let that put you off coming ...

Abstract for talk

A review of methods for specifying the target difference in randomised controlled trials (DELTA review)

When the sample size for a randomised controlled trial (RCT) is determined, a (target) difference is typically specified (following a Neyman-Pearson approach) which the RCT is designed to detect. This provides reassurance that the study will have the required statistical certainty and power. From both a scientific and ethical standpoint, selecting an appropriate target difference is crucial; too large or small a study is arguably unethical, wasteful and potentially misleading. Various methods exist to specifying a target difference, though their relative merits are uncertain. This project aimed to assess formal methods for specifying the target difference in a RCT sample size calculation. It consisted of two main elements a) Comprehensive systematic review based on electronic searches of a variety of biomedical and non-medical databases were performed to identify methods b) Two surveys of trialists were conducted to assess current knowledge and practice: one of the International Society for Clinical Trials and one of UK and Ireland based trialists. Seven methods for specifying the target difference were identified – anchor, distribution, health economic, opinion-seeking, pilot study, review of evidence base and standardised effect size; each with important variations. Some methods require strong assumptions to be used in the design of a RCT. The implications of each method from a statistical perspective will be highlighted. While no single method provides a perfect solution to a difficult problem, raising the standard of RCT sample size calculations and the corresponding reporting of them would aid clinical decision making and more efficient use of resources.
Our joint Highlands Local Group meeting with St. Andrews Unversity is taking place on the afternoon of Wednesday 21st November from 2pm to 5pm in the Maths Building (Lecture Theatre B), St. Andrews. The event involves three speakers from Duke University, The University of Shelffield, and Biomathematics and Statistics Scotland. Topics include modelling presence-only data as a point pattern with enviromental covariates, the implications for model error in statistical models, and the modelling of runoff sources in river catchments. Titles and abstracts which do the three talks alot more justice than this very brief summary are given at the end of this notice. The organiser (Janine Illian) would like a rough idea of how many people will attending for catering purposes so if you are planning to attend please send a message to ideally by Monday 12th of November. All, including non-members, are very welcome to attend this event. A programme for the event is also given at the end of this notice

Programme and abstracts for Joint Meeting at St. Andrews on 21st November 2012

2:00: Welcome
2:05 - 2:50: Tony O'Hagan,University of Sheffield: "Aspects of Model Uncertainty"
3:00 - 3:45: Mark Brewer, BIOSS, Aberdeen: "Source distribution modelling for compositional analysis in hydrology " 
3:45 - 4:15: coffee break
4:15 - 5:00: Alan Gelfand, Duke University, Durham, USA: "Point pattern modeling for degraded presence-only data over large regions"


Tony O'Hagan (University of Sheffield):

The MUCM project ( has been developing Bayesian statistical methods for quantifying and managing uncertainty in the outputs of complex mechanistic models. Such models are used in numerous fields to predict, understand and control complex physical systems, and are typically built out of differential equations that represent the best available scientific knowledge. For users of such models a key question is how accurate they are as predictors of the real-world system - this is the issue that MUCM addresses. One cause of inaccuracy is errors in the parameters that are input to the model, but another key factor is model discrepancy. "All models are wrong", and this is as true of mechanistic science-based models as it is of statistical models. Even when the correct, true values of input parameters are used, the model will not predict reality correctly, and the difference is model discrepancy. In this talk, I will introduce the methods developed in MUCM for tackling uncertainty in the predictions of mechanistic models. I will look particularly at the complications caused by model discrepancy, and will go on to consider the implications for model error in statistical models.

Mark J Brewer (Biomathematics and Statistics Scotland):
End-member mixing (EMM) is a method in hydrology for attempting to define the runoff sources in river catchments. It involves estimation of the relative proportions of water from different sources, and is often recorded as a time series. Given regular measurements of a chemical tracer on the target water body and, in addition, corresponding measurements for samples of known sources, it is possible to perform end-member mixing using compositional analysis taking a Bayesian random effects approach in a hierarchical framework, including covariates if appropriate. This talk considers the case where there are no separate data available for the source components, and develops a model for source distributions via nonlinear regression on the tracer/flow relationship and nonparametric density estimation. We allow these source component distributions to vary from year to year and apply the model to a data set from two streams in central Scotland, comprised of weekly or fortnightly readings of alkalinity over seventeen years. We conclude there is evidence of a change in source distribution over time; that corresponding to low flow conditions exhibits a gradual increase in alkalinity for both of two streams studied, whereas for high flow conditions alkalinity appeared to be rising for only one stream.

Alan Gelfand (Duke University):
Explaining species distribution using local environmental features is a long standing ecological problem. Often, available data is collected as a set of presence locations only thus precluding the possibility of a presence-absence analysis. We propose that it is natural to view presence-only data for a region as a point pattern over that region and to use local environmental features to explain the intensity driving this point pattern. This suggests hierarchical modeling, treating the presence data as a realization of a spatial point process whose intensity is governed by environmental covariates. Spatial dependence in the intensity surface is modeled with random effects involving a zero mean Gaussian process. Highly variable and typically sparse sampling effort as well as land transformation degrades the point pattern so we augment the model to capture these effects. The Cape Floristic Region (CFR) in South Africa provides a rich class with such species data. The potential, i.e., nondegraded presence surfaces over the entire area are of interest from a conservation and policy perspective. Our model assumes grid cell homogeneity of the intensity process where the region is divided into 37, 000 grid cells. To work with a Gaussian process over a very large number of cells we use predictive process approximation. Bias correction by adding a heteroscedasticerror component is implemented. The model was run for a number of different species. Model selection was investigated with regard to choice of environmental covariates. Also, comparison is made with the now popular Maxent approach, though the latter is much morelimited with regard to inference. In fact, inference such as investigation of species richness immediately follows from our modeling framework.
The next RSS Highland Local Group is organised jointly with the Epidemiology Research Group of SRUC. The meeting comprises a talk by Professor Peter Diggle (University of Lancaster & University of Liverpool) on the subject of 'Statistical Modelling for e-Health'. The abstract for the talk is given at the bottom of this email, just in case you missed the first announcement. The meeting is located in the relaxing grounds (well if the sun is shining) of SRUC, Stratherrick Rd., Inverness on Tuesday 7th May and will commence with refreshments at 3pm, with the talk starting at 3:30pm. We will also head-out for a sociable (but not too expensive) bite-to-eat in town after the event (restuarant located approximately 5-10 minutes walk from the railway station for those catching trains to Aberdeen), and every-one is welcome to join in this also. All are very welcome to the meeting, whether members of the RSS or not, so feel free to pass the word around. The organiser, Dr. Roger Humphry (, has requested that, if possible, you indicate your interest in attending the event to him beforehand so that he can plan for approximate numbers; although if you unexpectedly find you can attend, don't let that put you off coming.
Abstract for talk

Statistical Modelling for e-Health

There is a range of opinion on exactly what is meant by the term  e-health.  
In this talk I will take it to mean a body of methodology intended to extract 
useful information from health-related data that is routinely acquired, often 
in real-time, during the operation of a health care system, rather than through 
specific, planned research projects. My impression is that large, and rapidly 
increasing, amounts of such data are acquired and stored for possible future 
retrieval, but all-too-seldom analysed. Also, these data are often temporally 
and/or spatially referenced and can thereby be linked to other electronically 
accessible data-sets such as census records, consumer purchasing patterns or 
social media activity. 

In this talk, I will argue that current research on statistical modelling of 
spatially and/or temporally referenced data can make an important contribution 
to real-time analysis of e-health data, with a view to identifying underlying 
trends and, often more interestingly, unexpected departures from those trends. 
I will then describe several applications of these ideas that are in various 
stages of development, from completed projects to half-formulated ideas.

The next RSS Highland Local Group meeting is a popular talk on statistics at the Inverness Science and Technology Festival on Wednesday 19th June. The talk is being held at the Inverness College site on Longman Rd. at 7:30 pm.

It is 80th anniversary year of the Loch Ness Monster being first sighted in modern times (although I understand the first encounter was claimed for St. Columba just over 1,400 years previously). To celebrate this event our speaker, Dr. Charles Paxton (University of St. Andrews), will deliver a talk entitled "The Vital Statistics of the Loch Ness Monster" which is intended to inform by exploring what statistics can tell us about the probability of discovering new giant animals in freshwaters, as well as revealing the results of a statistical analysis of eyewitness testimony from Loch Ness. This talk represents the 2013 RSS Highland Local Groups contribution to getstats, the Royal Statistical Society's ten 10-year public understanding of statistics campaign.

There is no charge for this event, although it would help if you were able to book a place in advance by accessing (event located about half way down the page)

All are welcome.
It is a pleasure to announce that our next RSS Highland Local Group meeting is a joint meeting between our RSS Highlands Local Group and the University of St. Andrews held in St. Andrews on the afternoon (2pm) of Wednesday 27th November. The meeting, which will be held in lecture theatre C of the Mathematics Building of the University of St. Andrews (North Haugh), will include two talks with refreshments mid-way offering an opportunity for chat.

The first talk is provided by Professor Dave Woods (University of Southampton) who has expertise in experimental design selection and assessment and will be talking about the decision-theoretic approach to Bayesian design with a focus on a new criterion for design selection. An abstract for this talk which is titled 'Decision-theoretic design of experiments for model discrimination' is provided below.

The second talk is provided by Dr. Daniel Mortlock (Imperial College London) who has expertise in the problems of inference about the real world when made from incomplete or imperfect data. The talk will describe some schemes for formalising heuristic approaches to model testing within a Bayesian framework and will not assume a prior knowledge of astronomy or cosmology; this talk will provide useful and refreshing perspective to the biological or medical context of many of our meetings. An absract for this talk which is titled 'Bayesian Model Comparison in Astronomy and Cosmology' is provided below.

The full details of the meeting with abstracts are provided below.

Joint RSS St Andrews meeting St Andrews lecture theatre C, Maths building, North Haugh) : Wednesday 27th November 2013

2.00pm -2.05pm Welcome

2.05pm – 2.55pm Dave Woods (University of Southampton, UK) Decision-theoretic design of experiments for model discrimination

2.55pm – 3.30pm Coffee Break

3.30pm -4.20pm Daniel Mortlock (Imperial College, London, UK) Bayesian Model Comparison in Astronomy and Cosmology


Dave Woods (University of Southampton, UK)
Decision-theoretic design of experiments for model discrimination

The design of any experiment is implicitly Bayesian, with prior knowledge being used informally to aid decisions such as which factors to vary and the choice of plausible causal relationships between the factors and measured responses. Bayesian methods allow uncertainty in these decisions to be incorporated into design selection through prior distributions that encapsulate information available from scientific knowledge or previous experimentation. Further, a design may be explicitly tailored to the aim of the experiment through a decision-theoretic approach with an appropriate loss function. This talk will review the decision-theoretic approach to Bayesian design and then focus on a new criterion for design selection when the aim of the experiment is discrimination between rival statistical models. Motivated by an experiment from materials science, we consider the problem of early stage screening experimentation to choose an appropriate linear model, potentially including interactions, to describe the dependence of a response on a set of factors. We adopt an expected loss for model selection which is a weighted sum of posterior model probabilities and introduce the Penalised Model Discrepancy (PMD) criterion for design selection. The use of this criterion is explored through a variety of issues pertinent to screening experiments, including the choice of initial and follow-up designs and the robustness of design performance to prior information. Designs from the PMD criterion are compared with those from existing approaches through examples. Further issues, such as reducing the computational burden of the method for experiments with a large number of contending models, will be addressed if time allows. Some directions of current and future research will also be discussed.

Daniel Mortlock (Imperial College, London, UK)
Bayesian Model Comparison in Astronomy and Cosmology

Bayesian inference provides a self-consistent method of model comparison, provided that i) there are at least two models under consideration and ii) all the models in question have fully-specified and proper parameter priors. Unfortunately, these requirements are not always satisfied in real world problems. This is a particular difficulty in astronomy/cosmology: despite the existence of exquisitely-characterised measurements and quantitative physical models (i.e., sufficient to compute a believable likelihood), these models generally have parameters without well-motivated priors, making completely rigorous model comparison a formal impossibility. Still, huge advances have been made in cosmology in particular in the last few decades, implying that model comparison (and testing) is possible in practise even without fully specified priors. I review the above issues (without assuming any knowledge of astronomy or cosmology) and describe some schemes for formalising such heuristic approaches to model testing within a Bayesian framework.
Ther will be a RSS Highland Local Group meeting in Aberdeen on Wednesday 11th December. This will feature a talk by Dr David McLernon (University of Aberdeen) on the ‘The Development and Validation of Clinical Prediction Models' and will be held at 15:00 in Room 115 of the Health Sciences Building (Forresterhill campus, Aberdeen). This meeting will conclude with a short AGM which will provide an opportunity for all RSS and local group members to hold to account their committee and office holders and will include a discussion of our anticipated activities over the coming year.

The January RSS Highland Local Group talk by Professor Rosemary Bailey (University of St. Andrews)
titled From Rothamsted to Northwick Park: designing experiments to avoid bias and reduce variance
has now been rescheduled. It will be held a week later than originally anticipated on the 23rd January 2014
at 16:00 in Room 115 of the Health Sciences Building (Forresterhill campus, Aberdeen).
Tea, coffee and good company will be available prior (15:30 onwards) to the meeting.

It is a pleasure to announce that our next RSS Highlands Local Group meeting will take place on Friday 28th February at the Old Aberdeen campus of the University of Aberdeen. This is a joint meeting with the School of Biological Sciences, University of Aberdeen, and comprises two talks starting at 13:00 hrs.

The first talk will comprise a talk by Professor Simon Wood (Mathematical Sciences, University of Bath) entitled 'Simple statistical methods for complex ecological dynamics'. The talk will be in the Zoology Lecture Theatre of the Zoology Building off Tillydrone Avenue and will start at 13:00 hrs.

This will be followed by refreshments, and a short amble across the University Botanic Gardens to the Auris Lecture Theatre (at 23 St. Machar Drive) where we will welcome Dr. Roland Langrock (School of Mathematics and Statistics, University of St. Andrews) who will talk about his work on 'Nonparametric inference in latent-state models - with applications in ecology and finance'.

All are welcome to attend the talk whether RSS members or not, so feel free to pass the word around.

Notices regarding two of our upcoming HLG meetings

First the location for our necessary RSS Highlands Local Group Annual General Meeting (at 4pm on 19th November 2014) to be held in room 315 of the Health Science Building of the Forresterhill campus in Aberdeen. Any items for including in the agenda including nominations for committee members and the honorary secretary of the group can be sent to

Second (and statistically very much more interesting than an AGM) the announcement of our groups third statistical meeting of the year to be held as a joint meeting with the University of St. Andrews. This meeting, which will take place on the 26th November in Lecture Theatre C of the Mathematical Institute of the University of St. Andrews between 2-5 pm comprises two speakers, Professor Mark Girolami of the University of Warwick ('Quantifying pistemic uncertainty in ODE and PDE solutions using Gaussian measures and Feynman-Kac path integrals') and Dr Natalia Bochkina of the University of Edinburgh. See
Please find below details and Abstract for the RSS talk Monday 8th June to be given by Prof Richard Riley :

University of Aberdeen Foresterhill Site,
Polwarth Room1:029
3.30pm Tea and Coffee
4-5pm Talk

Title: Multivariate meta-analysis: advantages, disadvantages & applications

Richard Riley
Professor of Biostatistics
Keele University

Meta-analysis is the statistical synthesis of results from related primary studies, in order to quantify a particular effect of interest. It is an immensely popular tool in evidence based medicine, and is used to summarise treatment effects, to identify risk/prognostic factors, and to examine the accuracy of a diagnostic test or prediction model, amongst many other applications. Many primary studies have more than one outcome of interest, such as the treatment effect on both disease-free survival and overall survival, and researchers usually meta-analyse each outcome separately. However, such multiple outcomes are often correlated. For example, a patient’s time to recurrence of disease is generally associated with their time of death. By meta-analysing each outcome independently, researchers ignore this correlation and thus lose potentially valuable information. As well as multiple outcomes, other correlated measures may also be of interest for the meta-analyst such as multiple treatment effects (e.g. A vs B, and A vs C), and multiple performance measures (e.g. sensitivity and specific, calibration and discrimination, etc). In this talk, I describe how multivariate meta-analysis models can jointly analyse multiple effects and account for their correlation. I discuss the statistical advantages and disadvantages of the approach over standard univariate methods (which analyse each outcome separately). In particular, I illustrate the gain in precision and borrowing of strength that the correlation can bring, which makes more of the data we already have and reduces issues such as outcome reporting bias. Though the talk will detail statistical issues, it is intended for a broad audience and especially for those interested in systematic reviews and meta-analysis. Real examples (mostly from the medical field) will be used to illustrate the potential application of multivariate meta-analysis, including multiple outcomes, multiple treatments (network meta-analysis), test accuracy, and risk prediction.
Please find below the title and abstract for David Millers talk on the 16th July - Also note the venue is in the Zoology Building (ZG18) in Old Aberdeen. The time is confirmed at 3.30 for refreshments and 4pm start

David L Miller: Centre for Research into Ecological and Environmental Modelling University of St Andrews

Title: Recent advances in spatial modelling of distance sampling surveys

Abstract: Our understanding of a biological population can be greatly enhanced by modelling their distribution in space and as a function of environmental covariates. Such models can be used to investigate the relationships between distribution and environmental covariates as well as reliably estimate abundances and create maps of animal and plant distributions.

Here I'll give an overview of "density surface models", which consist of a spatial model of the abundance which has been corrected for uncertain detection via distance sampling methods. The spatial model consists of a generalised additive (mixed) model, which can include many varied components, such as smooth terms and random effects. In particular, I'll highlight: flexible detection functions, quantification of uncertainty in a two-stage model, correction for availability bias, alternative/unusual response distributions, autocorrelation and smoothing in areas with complex boundaries. I'll show how such models are easily constructed, fitted, checked and compared using the R packages dsm and Distance using an example study of black bears in Alaska.

Next meeting

Speakers: Prof Brian Cullis and Dr Alison Smith from the Centre for Bioinformatics and Biometric Research, Wollongong NSW, Australia
Title: Experimental designs for expensive multi-phase traits
Date: 28th September 2015
Time: 4-5 pm meeting with refreshments at 3.30pm
Venue: Suttie Centre; Room 402 (more details will follow closer to the time)

Multi-phase experiments are widely used in many areas of biological research. Key examples include experiments in agriculture, medicine and pharmaceutics where biological samples are assayed in order to compare responses of different treatments. In general, the treatments of interest are applied to experimental units in a first phase, but subsequent laboratory phases are required to measure the trait. In order to ensure valid inference and accurate prediction of treatment responses a sound experimental design, involving replication and randomisation in all phases, is required. Typically the laboratory phases are costly and there can often be insufficient material for full replication. In either case this necessitates a limit on the total number of samples that can be tested. In this talk we describe how statistically valid and cost-effective multi-phase designs can be achieved using the approach of Smith et. al. (2015). In terms of first phase replication, some treatments are tested in the laboratory as composite samples and some as individual replicate samples. Replication in the laboratory is achieved by splitting a relatively small number of first phase samples into sub-samples for separate processing. Model-based design techniques are used to obtain efficient designs for the laboratory phases, conditional upon the first phase design. The approach will be illustrated using an Australian wheat quality project that involved a series of field experiments in which a number of elite varieties were grown. The aim was to obtain accurate estimates of genetic effects, including variety by environment interaction, for a range of quality traits such as flour yield and dough and baking characteristics.

A.B. Smith, D.G. Butler, C.R. Cavanagh and B.R. Cullis. 2015. Multi-phase variety trials using both composite and individual replicate samples: a model-based design approach. The Journal of Agricultural Science, Vol 153, pp 1017–1029.

World Statistics Day Tuseday 20/10 RSS talk

Speakers: Dr Andrew Grey (Associate Professor, University of Auckland, New Zealand) and Fiona Stewart (University of Aberdeen)
Topic: Data Hunt! Outing a fraudster using statistics. Dr Andrew "Sherlock" Grey and Fiona "Watson" Stewart.
Date: 20th October 2015
Venue: MedChi Conference room, Polwarth building, University of Aberdeen, Foresterhill site
Time: 4-5pm (usual 3.30 for refreshments)

1) Dealing with doubts about data - a case study of a prolific author
Dr Andrew Grey, Associate Professor, University of Auckland, New Zealand
Andrew Grey, together with Mark Bolland, Ian Reid and Greg Gamble from the University of Auckland, have collaborated with Alison Avenell, University of Aberdeen, on systematic reviews of vitamin D supplementation. These methods threw up some unexpected questions about the data and research integrity. Andrew will talk about our investigation of one exceptionally prolific author, whose results were just too good to be true. He will discuss statistical and other techniques we used to examine the publications, and the ongoing saga of investigations by the journals involved.

2) Impact of such activities on the wider evidence
Fiona Stewart, University of Aberdeen, will also talk briefly about the methods used to assess the impact of this author’s publications on the wider evidence base, through citations in reviews and guidelines, and the potentially far-reaching consequences of questionable research findings.

DATE: Monday 29th April 2016
SPEAKER: Prof Paul Garthwaite, Professor of Statistics, Mathematics And Statistics, The Open University
TIME: 4-5pm (Refreshments as usual from 3.30pm)
VENUE: Health Science Building HSB 115

TITLE: ‘Tweaking’ variables to make them uncorrelated

Interpreting the results of a statistical analysis would be easier if all
variables were uncorrelated and all data sets consisted of orthogonal x-variables.
Principal components is one way of obtaining orthogonal/uncorrelated variables,
but the relationship between the principal components and the original x-variables
is complex. This talk proposes two transformations that give orthogonal/uncorrelated
components with a close one-to-one correspondence between the x-variables and
the components. The transformations have a number of useful applications including:
• a unified approach to the identification and diagnosis of collinearities
• evaluating the contributions of individual variables to a Mahalanobis distance
or the test statistic of a Hotelling test
• setting prior weights for Bayesian model averaging
• evaluating the contributions of individual variables to a multiple regression
• calculation of an upper bound for a multivariate Chebyshev inequality.
The talk will focus on the first three of these applications.

Next meeting:

Thursday 1st September, 2:30-5pm.
Lecture Theatre D, School of Mathematics and Statistics, St Andrews

A) Dr Andrew Parnell: Statistical palaeoclimate reconstruction: how fast can climate change?
B) Dr Christopher Ferro: Evaluating weather and climate forecasts

Please let Steve Buckland ( know if you would like to attend the dinner at Forgan’s (110 Market Street) following the meeting by 30th August.


2:30-3:30: Dr Andrew Parnell: Statistical palaeoclimate reconstruction: how fast can climate change?
Abstract: I will present the generic problem of reconstructing aspects of past climate based on proxy data from a statistical perspective. There are now numerous data sources to assist with such reconstructions, which all present their own advantages and issues. Whilst my group have developed one particular approach, based on Bayesian inversion of causal models of the climate-proxy relationship, we have currently only applied it to individual proxies (pollen) and individual sites. Our approach contrasts with much of traditional palaeoclimate reconstruction. I will point out the differences and explore some of the many possible extensions which require collaboration between climatologists, mathematicians, proxy specialists, and statisticians.

4:00-5:00: Dr Christopher Ferro: Evaluating weather and climate forecasts
Abstract: Probabilistic weather and climate forecasts are a key part of risk-based decision support, and assessments of forecast performance help to guide both our responses to forecasts and our development of forecasting systems. The definitive measures of forecast performance are proper scoring rules. We shall discuss the use of proper scoring rules in evaluating weather and climate forecasts and then introduce recent extensions that are suitable for measuring the performance of ensemble forecasts, and for measuring performance when the truth is uncertain.

Richard Sylvester, well known for his work with EORTC (European Organisation for Research and Treatment of Cancer) especially urology, will be in Aberdeen on 9th May and has agreed to do a RSS talk while here. The time has to be a bit earlier than normal 2.15 Tea for a 2.30 start just so that we are able to take the opportunity of his visit:

DATE: Tuesday 9th May 2017
SPEAKER: Dr Richard Sylvester, EORTC (semi-retired)
TIME: 2.30-3.30 the actual talk (Refreshments from 2.15pm)
VENUE: Health Science Building HSB 115

TITLE: "Assessment of patient-level predictive factors in IPD meta-analyses"

It would be based on an actual IPD meta-analysis of treatment effect in bladder cancer and show both the "wrong" and the "right" way of determining treatment/covariate interactions, separating the within and between trial interactions using Stata.

DATE: Wednesday 28th June 2017
SPEAKER: Prof Elizabeth Thompson, University of Washington
TIME: 16:00
VENUE: Macaulay B Lecture Theatre, James Hutton Institute, Aberdeen
TITLE: "Mapping causal DNA through the shared descent of genome in population samples"

The relatedness of individuals is reflected in the close similarity of their DNA that is descended from shared common ancestors. Genome descends and functions in segments; models and methods for the detection of shared segments of genome can greatly increase accuracy. Modeling the dependence in the DNA both among individuals and across genome locations is key to using modern genomic data in the mapping the locations of DNA that contribute to a quantitative trait. With modern genetic marker (SNP) data, we can estimate pairwise proportions of genome shared both at specific genome locations and globally across the genome, without the need for prior information on pedigree relationships. We present models for using estimated genome sharing in gene mapping, and also show how Kullback-Leibler information can be used to provide a more reliable assessment of the meaning of a linkage signal in ascertained samples. Additionally, genome descends jointly to current members of a population: pairwise analyses lose information. I will present some recent approaches to estimation of the joint patterns of shared genome segments among multiple individuals and indicate how this additional information can be important in gene mapping.
We are delighted that Bendix Carstensen a senior statistician from the Department of Biostatistics, University of Copenhagen will be visiting us at the University of Aberdeen for 2 days in August. During this time there will be a series of talks on methods of Survival and a workshop as an introduction to the Lexis-machinery in the Epi-package for R, allowing users to get to grips with parametric survival and multistate models.
One of the talks has been detailed to be a RSS Local Highland Group meeting talk

DATE: Thursday 17th August 2017
SPEAKER: Bendix Carstensen, senior statistician, Depart of Biostatistics, University of Copenhagen
TIME: 16:00 (tea at 3.30 pm)
VENUE: IMS Level 7, Forresterhill, U of A, Aberdeen
TITLE: The resurrection of time as a continuous concept in biostatistics, demography and epidemiology

In classical demography the actuarial estimator of the survival function has been used for centuries, typically with one year time intervals. This estimator was also the default in medical science and epidemiology till the advent of computers, when the Kaplan-Meier estimator and later the Cox-model became the de facto standard for analysis of survival data and more generally, for follow-up data from cohort studies. The reporting of survival curves in the medical literature is therefore almost exclusively as the well-known noisy step curves, which is then left to eye-ball smoothing by the reader.
In epidemiology there has also been a tradition in analysis of occurrence rates by tabulation of events and risk time in five- or ten-year intervals, and fitting models with a separate parameter in each interval. In some circumstances such as Age-Period-Cohort models based on data classified in Lexis triangles this approach gives distinctly illogical results [1].
Both of these approaches are essentially based on models where the effect of time is taken to be different in different intervals, and the intervals taken as being exchangeable | the inherent ordering of time-intervals are ignored in the model. For the very broad time-intervals the usual problems associated with categorization applies [2]. Only in the reporting of the effects is the ordering of time-intervals reintroduced, either by connecting the estimates from broad time-categories, or hiding the ragged nature of the estimates by only showing cumulative effects.
I will advocate the use of models using the quantitative nature of time in modeling of occurrence rates, by imposing restrictions on the time-effects re ecting this. This can be seen as a combination of the very ne subdivision of time used in the non-parametric modeling and assumptions about continuous, smooth effects of time. Besides the ability to show the time-effects directly on the rate-scale, this approach also includes the possibility to accommodate multiple time scales such as age and duration of disease simultaneously. It also readily extends to estimation of transition rates in multistate models [3]. Parametric estimates of transition intensities also allow derivation of demographic quantities such as residual life time and years of life lost.
An essential prerequisite for practical handling of these models is a proper way of data representation. I shall describe the philosophy of the Lexis machinery for representation of multistate data on multiple time scales implemented in the Epi package for R[4, 5]. Furthermore I will show how otherwise intractable quantities from large complex multistate models can be handled by simulation [6], even if models also involve time since entry to intermediate states as timescale. Examples will be drawn from the clinical literature, and I shall indicate directions for further work on devising proper measures of uncertainty of such quantities.

[1] B Carstensen. Age-Period-Cohort models for the Lexis diagram. Statistics in Medicine, 26(15):3018{3045, July 2007.
[2] Frank Harrell. Problems caused by categorizing continuous variables., 2015.
[3] S. Iacobelli and B. Carstensen. Multiple time scales in multi-state models. Stat Med, 32(30):5315{5327, Dec 2013.
[4] Martyn Plummer and Bendix Carstensen. Lexis: An R class for epidemiological studies with long-term follow-up. Journal of Statistical Software, 38(5):1{12, 1 2011.
[5] Bendix Carstensen and Martyn Plummer. Using Lexis objects for multi-state models in R. Journal of Statistical Software, 38(6):1{18, 1 2011.
[6] Bendix Carstensen. Simulation of multistate models with multiple timescales: simLexis in the Epi package. or Epi vignette, 2015.

Date: Friday 13th October; 4pm (refreshments as usual from 3.30)

Venue: UoA, Forresterhill Polwarth Building room 1.029

Speaker: Rebecca Walwyn, Principal Statistician, Clinical Trials Research Unit, Leeds Institute of Clinical Trial Research, University of Leeds. Rebecca has expertise in the design and analysis of complex intervention trials.

Title: Beyond standard factorial designs: Building on design of experiments methodology for clinical trials of complex interventions.

Abstract: Complex healthcare interventions, such as psychotherapy and surgery, are often defined as interventions that contain several potentially interacting components. Collins and colleagues have proposed that factorial trial designs are used to estimate the individual and combined effects of components of complex interventions. Fisher claimed that one of the benefits of a factorial design is it’s ‘wider inductive basis’, enabling factors to be studied under a variety of conditions. A potential barrier to the uptake of Collins’ proposal is the recommendation that factorial designs are only used in clinical trials when it is safe to assume that there will be no interactions or when the trial is powered to detect realistic interactions. Collins went on to propose a multiphase optimisation strategy (MOST), following design of experiments (DoE) literature from the 1970s and 80s. Optimisation of the intervention package is carried out in two phases: screening and refinement; optimisation follows theoretical development and requires confirmation. Building on Collins’ MOST, I will outline a refined research strategy for developing and evaluating novel and widely adopted complex interventions. I will then illustrate how the four features of intervention complexity highlighted by the MRC could be systematically mapped onto generic experimental designs, each of which can be viewed as a specific extension of a standard factorial design. I will then focus on what I term multilevel and cascading interventions.

There will be a RSS joint Highland Local Group and Environmental Statistics Section meeting ‘Analysing Environmental Data’ on Monday 23rd October 2017 between 1:15 pm and 4:30 pm. The meeting will be held at Scottish Natural Heritage, Great Glen House, Leachkin Road, INVERNESS, IV3 8NW.

This will provide an opportunity to present and discuss progressing or recently completed statistical analyses of physical, biological, terrestrial, marine and atmospheric environmental data. Analysis of such data is often challenging and by sharing experiences and expertise all can learn, to help progress analyses, and refine interpretations.

The first session will be presentations of four 15 minute talks on physical or biological environmental statistical or modelling analyses from throughout the north of Scotland (vaguely defined as anywhere north of and including St. Andrews). Each talk will include a brief discussion opportunity. The meeting organisers therefore invite abstracts for consideration for this part of the meeting. The abstract need not describe a definitive analysis although it would be appropriate if results were available and that at least some conclusions had been reached. The purpose of these talks is not to ‘show-off’ the work, but to generate discussion from which we will all learn and thereby improve our current and future analyses. The meeting organisers will consider these and expect to assemble a balanced selection of talks including the range of submitting institutes. The abstract should include the name of the speaker, title, institute, a 200 word summary, and a contact email address. Abstracts should be sent to and received before 8:00 am on Monday 11th September. A decision on the presentations selected will be sent to all those proffering an abstract by 8:00 am 18th September.

The second part of the meeting comprises an invited talk by Dr Adam Butler. Dr Butler is a senior statistician at Biomathematics and Statistics Scotland in Edinburgh with extensive experience of analysing environmental data and his talk (title not yet available) will be based on his experiences of analysing environmental data.

This meeting is held under the auspices of the Royal Statistical Society’s Highland Local Group and also Environmental Statistics Section, and is being organised by Megan Towers (SNH), Roger Humphry (SRUC) and Malcolm Hall (Scottish Government). There is no charge to attend the meeting although all attendees (other than Dr Butler) will be required to meet their own travel and subsistence costs. Tea and coffee will be available at the meeting, and while there is no charge for this, a small donation of your choosing to help cover these costs (if you are not a member of the Royal Statistical Society or giving a talk), would be gratefully received. However we do not wish to embarrass anyone, all will be welcome to refreshments, and if you are unable to donate don’t let it stop you attending.

Due to the size of our venue, kindly made available to us by SNH, the meeting is limited to 60 participants (including speakers). It would help if you could send a note letting us know that you intend to attend by the 17th October to This will ensure that space will be kept for you in the unlikely event that we are oversubscribed, and will also help us to order an appropriate number of refreshments. But if you don’t manage to send this note, you are welcome to take a chance by simply turning up.

Submission of abstracts: by 8:00 am 11th September
Decisions on abstract: by 8:00 am on 18 September
Requested registration: by 8:00 am on 17th October
Meeting date and times: 1:15-4:30 23rd October 2017

November meeting:

Phil Crook (Secretary RSS International Development Working Group [IDWG]) has arranged to come up to Aberdeen to hold an information session, ultimately to talk about opportunities for the IDWG. They are particularly interested in some Health and Climate change statistics information (and possible collaborations). This is going to take a slightly different format to our usual meetings. There’ll be a 1 to 1½ hours talk to include questions and answers.

Title: “Measuring Sustainable Development”
Speaker: Phil Crook (Secretary RSS International Development Working Group [IDWG])
Venue: Foresterhill Health Sciences Building, Room 115, Aberdeen
Date: November 9th 2017
Time: 3-5pm with Tea

Let us know if you would like to attend using Skype if you can’t make it to Aberdeen

In September 2015 the United Nations General Assembly adopted seventeen aspirational "Global Goals”, stretching from “No Poverty” and “Zero Hunger” through “Gender Equality” and “Decent work” to “Climate Action” and “Peace, Justice and Strong institutions”.

The goals are associated with 169 targets and 241 global indicators and the annual report on progress towards the targets will cover every country, developed as well as developing. Data disaggregation to ensure that no-one is left behind will be a key feature of reporting. Experience in the UK and globally shows that targets and indicators such as these can have profound effects.

The talk will cover the process and politics leading up to global goals, the experience with the Millennium Development Goals which preceded the SDGs, and then the SDGs themselves and the statistical challenges and opportunities they are throwing up.

Dear All

Here are details about the talks planned for our next Joint Meeting with St Andrews


Wednesday, 18th of April 2018 - Joint RSS Highland group- St Andrews meeting.


Venue: Lecture Theatre C, Physics building, North Haugh.




2.00pm -2.05pm                      Welcome


2.05pm – 2.55pm                    Marta Blangiardo (Imperial)

2.55pm – 3.30pm                    Coffee Break


3.30pm - 4.20pm                     Thordis L. Thorarinsdottir (Norwegian Computing Centre, Oslo)




Speaker’s biographies


Marta Blangiardo is a senior lecturer in Biostatistics at the Department of Epidemiology and Biostatistics, Imperial College London (UK) affiliated with the MRC-PHE Centre for Environment and Health (  Her main research interest is the development of Bayesian hierarchical models, especially in relation to the use of multiple data sources, with application in a wide range of areas, including genetics, environmental science, and epidemiology. Recently she has been working on spatio-temporal epidemiological surveillance for chronic diseases, confounder adjustment in ecological studies of environmental epidemiology and models for multi-pollutant exposure in a time series framework.



Title: A data integration approach for improving inference in area-referenced environmental health studies (joint work with Monica Pirani, Alexina Mason, Anna Hansell, and Sylvia Richardson)

Abstract: Study designs where data are aggregated into geographical areas are extremely popular in environmental epidemiology. These studies are commonly based on administrative databases and, providing a complete spatial coverage, are particularly appealing to make inference on the entire population. The ecological nature of these studies, however, does not allow the direct inclusion of individual-level risk factor data. In the presence of unmeasured potential confounding factors, risk effect estimates are prone to bias. Here, we show how to improve inference drawn from area-referenced environmental health-effect studies, proposing a Bayesian approach that augments measured area-level covariates with an ecological propensity score estimated upon individual-level data from sample surveys. This scalar index acts as a proxy for the unmeasured ecological confounders and represents a useful tool for overcoming the problem due to the incomplete spatial coverage of the individual-level data. In contrast to the main literature on propensity score for confounding adjustment where the exposure of interest is confined to a binary domain, we generalize its use to cope with ecological studies characterised by a continuous exposure. The approach is illustrated using simulated examples and a real application investigating the risk of lung cancer mortality associated to nitrogen dioxide in England (UK).


Thordis L. Thorarinsdottir is a statistician working within the fields of spatial and space-time statistics, Bayesian methods and forecasting. She is a Chief Research Scientist in the Statistical Analysis, Machine Learning, and Image Analysis (SAMBA) group at the Norwegian Computing Center in Oslo, Norway. Her current work focuses on environmental applications to solve problems on uncertainty quantification, probabilistic prediction and model evaluation in collaborations with climate ecologists, atmospheric scientists and hydrologists.


Title: Does Bayes beat squinting? Bayesian modelling of cluster point process models

Abstract: Point process data arises naturally in various fields of science such as biology, ecology, epidemiology, and environmental sciences. However, the point process modelling framework is very involved and inference can often only be performed approximately and with great care. At the same time, a great number of different models are available where the subtle differences between the individual models can be hard to detect. In this talk, we discuss to which extend Bayesian modelling approaches may be applied to the class of cluster process models.  Cluster point processes have the following general structure. There is a point process of cluster centers and to each cluster center is associated a random number of points forming a subsidiary process, where the points in the subsidiary process are distributed about the cluster center in some specific way. A model for a cluster point process thus consists of three components; a component describing the cluster center process, a component describing the cluster sizes, and a component describing the distribution of the subsidiary points around the cluster center, the dispersion process. We consider how Bayesian approaches may be used to perform inferences for all three components, a feat which is often not possible using other inference methods

We are delighted to announce that Professor Deborah Ashby, Chair in Medical Statistics and Clinical Trials at Imperial College London (and RSS President in waiting), will be visiting us at the University of Aberdeen in April. Deborah has kindly agreed to deliver a talk for the RSS Highlands Local Group. The details are below and the title will be announced in due course.

Date: Wednesday 11th April 2018

Speaker: Deborah Ashby, Chair in Medical Statistics and Clinical Trials at Imperial College London

Time: 3pm for 3.30pm start

Venue: Room 115, Health Sciences Building, Foresterhill, UoA, Aberdeen

Title: Better benefit-risk decision-making in the regulation of medicines: new opportunities for statistical and data sciences

Abstract: Until recently, assessment of the benefit risk balance for a medicine has been entirely informal, but there is now growing interest among drug regulators and pharmaceutical companies in the possibilities of more formal approaches to benefit-risk decision-making, including those which explicitly take patient perspectives into account. Pharmacoepidemiological Research on Outcomes of Therapeutics by a European Consortium (PROTECT) was a project funded under the Innovative Medicines Initiative as a collaboration between academic, pharmaceutical, regulatory and patient organizations. Based on work from PROTECT we review current methodological approaches, and illustrate them with the case-studies where benefit-risk is finely balanced. We will introduce the PROTECT Benefit-Risk Roadmap, designed to help those wishing to find their way through this evolving arena (, and highlight recent statistical developments and current challenges.

Dr Jeff Ralph, Royal Statistical Society William Guy Schools’ Lecturer and a visiting academic at the University of Southampton, will be delivering a talk for the RSS Highlands Local Group on Wednesday 13th June. Details and abstract below.

Date: Wednesday 13th June 2018

Speaker: Jeff Ralph, Royal Statistical Society William Guy Schools’ Lecturer and a visiting academic at the University of Southampton

Time: 3.30pm for 4pm start

Venue: Room 115, Health Sciences Building, Foresterhill, UoA, Aberdeen


The History of Inflation Measurement

Consumer price inflation figures from the Office for National Statistics are among the most influential of all Official Statistics. They are used for a wide variety of important purposes from indicating the health of the economy to the adjustment of pensions and benefits. Behind the numbers sits a sophisticated methodology with a long history of development.

This talk describes the development of consumer price inflation measures from their origins at the start of the 18th Century to the current day. It identifies some of the visionary individuals who contributed to establishing the foundations, the dates when important changes were made and the social and political factors that drove the developments.

About Jeff ...

Dr Jeff Ralph worked for the Office for National Statistics for 14 years focussing mainly on price statistics. He is the current Royal Statistical Society William Guy Schools’ Lecturer and a visiting academic at the University of Southampton. Jeff was a member of the Technical Advisory Panel for Consumer Prices for two years and he is joint author of two books - A Practical Introduction to Index Numbers, published by Wiley in 2015 and Inflation: History and Measurement, published by Palgrave Macmillan in 2017.

Please see below the details of two joint RSS HLG - UoA SBS talks that will take place on the afternoon of Weds 27th June 2018. The speakers are the co-authors of the book "Bayesian population analysis using WinBUGS – a hierarchical perspective" and we are grateful that they have agreed to give these talks in conjunction with their workshop on "Integrated population modeling" organised by former Aberdeen researcher Chris Sutherland in advance of the International Statistical Ecology Conference at St Andrews the following week.

Tea / coffee in foyer of the Zoology Building, Tillydrone Avenue, Aberdeen.

Marc Kéry, Swiss Ornithological Institute, Sempach, Switzerland
Use of occupancy modelling in bird population studies: imputing missing data, confronting preferential sampling and modelling dynamic rates

Michael Schaub, Swiss Ornithological Institute, Sempach, Switzerland
Inference about population processes by combining counts and demographic data using integrated population models

Date: Monday 22nd Oct 2018

Speaker: Matteo Quartagno, Research Fellow, Dept of Medical Statistics, Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine

Time: 3pm for 3.30pm start

Venue: Room 115, Health Sciences Building, Foresterhill, UoA, Aberdeen

Title: Handling Missing Multilevel Data with Joint Modelling Multiple Imputation

Abstract: Abstract: Multiple Imputation (MI) is a flexible tool to handle missing data that has been increasingly used in recent years. It broadly consists in filling in the missing values multiple times, creating several completed datasets; these are then analysed with standard techniques obtaining different parameter estimates, that are finally combined with Rubin’s rules. One of the conditions for the validity of MI is that the two models used for (i) imputing and (ii) analysing the data need to be compatible. For this reason, when the partially observed data have a multilevel structure, both models need to reflect this. In this talk I am going to present an imputation technique, known as Joint Modelling imputation, based on running an MCMC sampler after defining a joint imputation model for the partially observed variables. This imputation strategy is particularly appealing for imputing data compatibly with a multilevel analysis model and is implemented in the R package jomo. I will explain under which circumstances simple JM imputation works properly, and I will explore possible solutions to situations where it doesn’t. Finally I will conclude by outlining plans for future research in this area.

We are delighted to announce our first RSS Highlands Local Group talk of 2019. Rik van Eekelen, PhD Candidate from the Centre for Reproductive Medicine, Academic Medical Centre, Amsterdam, will deliver a talk on Wed 9th January at 2.30pm. This will be followed by our AGM, which anyone can attend. The details are below:

Date: Tues 15th January 2019
Time: Refreshments from 2.30pm with talk starting at 3pm
Venue: Room 1:029, Polwarth Building, Foresterhill, University of Aberdeen

Please see Rik’s abstract below along with the agenda for the AGM.

3pm Guest speaker: Rik van Eekelen, PhD Candidate, Centre for Reproductive Medicine, Academic Medical Centre, Amsterdam

The Cox proportional hazards model for treatment effects in observational data: possibilities and pitfalls Couples who are unable to conceive after twelve months of trying, and in whom no barriers to conception can be found, are said to have unexplained subfertility. It remains uncertain if these couples should receive in vitro fertilisation (IVF). We compared one cohort that followed couples from their first IVF cycle to two cohorts that followed couples who continued trying to conceive naturally. Three methodological and statistical issues arose during analysis. First, we may ask ourselves: what is the epidemiological framework of the research we are conducting? Could it be considered prediction of a counterfactual? Second, the treatment effect was non-proportional over time. If the event probabilities for both groups are of primary interest, what are our options? Third, we had concerns that the three cohorts were not comparable due to possible selection bias. We showcase the nature of this issue in data on insemination treatment.