Research in the Department of Statistics
Description of Current Research Activity Areas
Biostatistics and novel regression methodologies
Statisticians at Auckland are developing new statistical methods for a whole range of problems, many of which originally came to the notice of the statistical community in a medical or biostatistical context, but whose solutions are generally applicable. Much of this work relates to forms of regression, i.e. to finding novel ways to understand or predict the behaviour of variables of compelling interest using information gathered on other variables. Examples include models and generalised estimating equation methods to better understand correlated binary data, improved methods for longitudinal data analysis, using smoothing techniques to reveal relationships in multivariate data without making restrictive assumptions, adapting regression methods to allow novel approaches to sampling that can increase the efficiency of observational studies (e.g., generalisations of case-control sampling). Several researchers are also actively involved in practical medical research.
Statistical computing
Standard commercial statistical packages are good for routine analyses of sets of data without non-standard features. Such data sets are rarer than you might think. Commercial packages are fast and implement algorithms and procedures that were seen to be viable at the time the package was written. For statistical research one requires more flexibility than these packages generally provide. A statistician has an idea for a non-standard approach to the problem and wants to try it out. In these situations a more flexible system is required. It is not the time it takes the computer to execute the commands that is important but rather the time it takes the statistician to develop new software. Ongoing research and software development in the Statistics Department is building computing environments that increase the effectiveness and efficiency of the statistician. There is associated research on statistical graphics, on developing software to implement specific new statistical methods and on computer-intensive statistical methods such as Markov Chain Monte Carlo.
The analysis of survey data
Over the years, researchers have been analysing data from complex surveys as if it came from random samples. The results of such analyses can be extremely misleading. Standard methods have to be adapted to allow for the special features of survey data. Alastair Scott has been a world authority in this area for many years. Research is also being done into novel ways of obtaining and analysing market research data, multilevel modelling, the effects of interviewer variability on subsequent analyses, and the design of panel studies.
Statistical ecology and the sampling of biological populations
Auckland is known for its research into the estimation of animal numbers. George Seber is a world authority and the author of the standard textbook on the subject. The way that animal and fish populations cluster make for special sampling problems and allow for novel approaches. There is an active research programme in adaptive sampling (which involves sampling most heavily where you have found something) and the first full-length book on the subject by George Seber and Steve Thompson (Steve was also a member of our Department for several years) was published in May, 1996.
Brian McArdle is an ecologist turned statistician with wide experience of all branches of ecological statistics. His particular interests are in the analysis of community structure and change (i.e. multivariate statistics) and in applications of statistical methods in population ecology.
Forensic statistics
Forensic science raises a host of statistical problems relating to the interpretation of physical evidence including DNA evidence and other aspects of genetics. Much forensic work has been carried out on a very ad hoc basis and needs to be put on a more solid scientific footing. We have an active research programme in this area in collaboration with ESR (the government forensic service). Several researchers in this programme won the 1990 award for the best paper in the Journal of the (international) Forensic Science Society.
Statistics Education
Students preconceptions about statistical concepts such as randomness are deep seated, influence their ability to learn, and are extremely hard to change. Students’ statistical intuitions and conceptions, how they learn, how they reason, and how statistical education can be changed to take advantage of such knowledge is a growing area of research. Quality management as a means of achieving continuous improvement in a teaching environment is an associated research area being developed here.
Experimental design and quality improvement
The design and analysis of experiments is still a fertile research area. It has received renewed impetus because the emphasis quality improvement gives to using experimentation to obtain good quality data to make improvements, in areas where obtaining data is often extremely expensive. Improvements in computation have also allowed increased flexibility in choice of design and analysis. We have several researchers working in this area.
Stochastic Processes
Researchers are working in Applied probability, Markov chains, queuing theory, stochastic networks, stochastic control, filtering, prediction and stochastic differential equations and a wide range of applications including telecommunication systems, biological modelling, epidemiology, financial mathematics and dynamical systems. Collaborative research involves agricultural research establishments and telecommunication companies.
Bayesian statistics
Several members of the department do Bayesian modeling. Present research includes implementation of sophisticated state-space dynamic models, mixed effects models and multiple event failure time models. The latter is funded by a Marsden grant and also includes development and assessment of techniques for model evaluation.
Fisheries modeling
The department has strong links with the Leigh Marine Laboratory and has expertize in most areas of fisheries modeling. The strength of this research comes from the ability to develop and apply new and novel models to fisheries data, little of which is amenable to "standard" analyses. Research includes estimating mixing proportions of mixed stock fisheries, modeling environmental effects on growth and abundance of fish using nonlinear and generalized linear mixed effects models, estimation of size-selectivity curves, and application of Bayesian state-space methodology for stock assessment and stock-recruitment modeling.
Current Research Areas of Staff
Sampling and the analysis of survey data
Alastair Scott, George Seber, Peter Danaher
Estimation of animal abundance and adaptive sampling
George Seber, Russell Millar
Longitudinal data and correlated responses
Alastair Scott, Alan Lee, Robert Gentleman, Chris Wild, Thomas Yee
Nonparametric Regression
Alan Lee, Chris Wild, Robert Gentleman, Thomas Yee
Case-control sampling and extensions
Alastair Scott, Chris Wild, Alan Lee
Statistical Computing and the "R" statistical programming language
Ross Ihaka, Robert Gentleman, Paul Murrell, Chris Triggs, David Scott
Medical Statistics
Patricia Metcalf, Renate Meyer, Alastair Scott, Chris Wild, Robert Gentleman, George Seber, Alan Lee, David Scott, Roger Marshall, Thomas Yee
Forensic Statistics
Chris Triggs, John Buckleton
Nonparametric methods
Alan Lee
Operations Research
Ilze Ziedins, Wiremu Solomon, Geoffrey Pritchard, David Ryan
Experimental design
Chris Triggs, Arden Miller
Quality Improvement and Industrial Statistics
Arden Miller, Peter Mullins, Chris Wild, Robert Gentleman
Multivariate Analysis
Renate Meyer, Chris Triggs, Brian McArdle, Marti Anderson
Statistical Education
Constance M. Brown, Maxine Pfannkuch, Chris Wild, Lynne Gilmore, Matt Regan
Stochastic Processes / Applied Probability
Ilze Ziedins, David Scott, Geoffrey Pritchard, Wiremu Solomon
Time Series Analysis
Constance Brown, Ross Ihaka, Alan Lee
Analysis of circular and spherical data
Alan Lee
Markov chain Monte Carlo methods
Renate Meyer, David Scott, Geoff Nicholls, Ilze Ziedins, Wiremu Solomon
Fisheries
(gear selectivity, fish growth, sequential population analysis)
Russell Millar
Philosophy of Statistics
(e.g. the role of conditioning)
Russell Millar
Statistical Ecology
Russell Millar, Brian McArdle, George Seber, Marti Anderson
Multivariate Modelling and Inference
Brian McArdle
Fitting Statistical Models to Population Density Data
Brian McArdle
Consulting Service
The Department operate a Statistical Consulting Service which provides statistical services on either a fee per service or retainer basis to internal and external clients. The Service can call on several experienced consultants with a wide range of experience and expertise. Services provided include
- Advice on design of studies
- All types of data analysis
- Expert witness service
- Survey design and analysis
- Advice on statistical computing packages
Enquiries are welcome at any time. The Director of the Consulting Service is Associate Professor Chris Triggs. He can be contacted on email triggs@stat.auckland.ac.nz, fax 64-9-373-7018, telephone 64-9-373-7599 Extn 8750. Alternatively you may contact the head of department, Associate Professor Alan Lee email lee@stat.auckland.ac.nz, fax 64-9-373-7018, telephone 64-9-373-7599 Extn 8749, or the departmental manager Kathy Henry, email kathy@.stat.auckland.ac.nz, fax 64-9-373-7018, telephone 64-9-373-7599 Extn 8741.