For general information regarding scholarships, please see the Scholarships and Research Grants webpage.
These Summer Scholarships are available to suitable undergraduate students.
Closing date for applications: Friday 27th August 2010 (4.00 pm)
Aim: To give students research experience and acquaint them with the research activities of the Department and with potential graduate supervisors.
Emolument: $5,000 non-taxable
Duration: Full time employment for eight to twelve weeks (400 hours)
A student is required to input data, create summary statistics and evaluate trends related to a survey of cardiology-related medical imaging within New Zealand. The first survey was completed in 2005, and we propose to repeat this in December 2010. This is a combined project between the Department of Statistics and the Department of Medicine and would be of interest to someone interested in pursuing a career in medical statistics.
Contact: Prof Chris Triggs cm.triggs@auckland.ac.nz, Katrina Poppe k.poppe@auckland.ac.nz, Dr Gillian Whalley
We are looking for a student with an interest in statistical graphics. Using a dataset of normal values, you will be developing a package that allows dynamic display of individual measurements against population series. The aim is to create an integrated program for detecting abnormalities in heart size derived from ultrasound measurements. This is a combined project between the Department of Statistics and the Department of Medicine.
Contact: Prof Chris Triggs cm.triggs@auckland.ac.nz, Katrina Poppe k.poppe@auckland.ac.nz, Dr Gillian Whalley
The aim of this project is to assist in developing a model of changes in family structure and socioeconomic factors that can be used to predict health outcomes and needs for pre-school children. Longitudinal data is available for over 1000 children and their families, from the birth of the child to the end of their 5th year. Regression and Markov chain models will be used to build a dynamic micro-simulation of family status and health outcomes. The project is joint between the Statistics Department and COMPASS (Centre of Methods and Policy Application in the Social Sciences). Supervisors will be Alan Lee, Ilze Ziedins, Roy Lay-Yee and Janet Pearson.
(Please also contact us if you are interested in applying for an HRC summer scholarship, the closing date for which is 15 September.)
Contact: Ilze Ziedins i.ziedins@auckland.ac.nz
Wave-cut platforms are a characteristic feature of the Hauraki Gulf and support a unique assemblage of intertidal reef species. These often extensive platforms are typically raised <1 m above the mean low water mark and consequently are severely threatened by rising sea level. This project will assess the relationships between species distributions and height on the shore and assess how overall assemblages will be affected by rising sea level. This will involve univariate and multivariate analysis of existing datasets, but also additional surveys of intertidal reefs around the Auckland region as part of an Auckland Regional Council rocky reef monitoring program.
This project will best suit a student with some biological background who is willing to be involved in surveys of intertidal reef communities over the summer break.
Contact: Nick Shears n.shears@auckland.ac.nz
The evolution of a set of species is generally represented by a tree or phylogeny. Such tree can be estimated from molecular sequences. A wide range of methods are available for that purpose, ranging from slow and accurate ones to fast and approximate approaches. The goal of this project is to examine the differences between tree reconstruction methods using multi-dimensional scaling. This dimension reduction technique will help us characterize the main features of the estimated trees that explain the differences between methods. This project will main involve writing short scripts in Perl (Python or even C/C++ or Java would do as well) and R.
Contact: Stéphane Guindon, s.guindon@auckland.ac.nz
Screening experiment are used to select a model from a set of candidate model. The ability of a design to discriminate between candidate models depends on the mathematical properties of the design matrix. One key characteristic that directly impacts this ability are the linear dependencies between columns of the design matrix. This project, will look at the model discrimination ability of 2-levels designs that have been created to have optimal properties in terms of their ability to discriminate between candidate models. The project will involve writing programmes in R that can be used to evaluate and rank such designs.
Contact: Arden Miller, a.miller@auckland.ac.nz
What are the actual effects of many of the so-called “risk factors” to which we are commonly exposed? How large are the risks of smoking, cholesterol, alcohol consumption, triglycerides, obseity, saturated fat, and failing to brush teeth twice a day? What, if any, are the interactive effects that are known?
Conversely, a large number of strategies for reducing risk of disease have been promoted – do they work? And to what extent, in absolute terms? How confident can we be that these effects are real?
Many hundreds of scientific papers have been written regarding these effects, and there have been several meta-analyses that synthesise and summarise these results. Generally, these results have not been put together in a comprehensive way designed for popular consumption. This is the objective of the proposed study.
There are many hundreds of papers published regarding these risks, and these interventions (among others). By focusing on review papers, we can reduce the work to be done here to review of perhaps as few as 20 papers.
The project is essentially a meta-analysis of several meta-analyses. It is not considered appropriate at present to prepare for submission to a scientific journal, so the need for statistical rigour is consequently reduced somewhat. Summary tables and simple weighted averages to get point estimates will in general be sufficient for our needs.
Contact: Peter Mullins, pr.mullins@auckland.ac.nz
This project is a continuation of work with the Cardiovascular Unit at Auckland City Hospital. A simulation and optimization module has already been constructed, and work will continue on improving the optimization module, and analyzing new data from the Cardiovascular Unit and, possibly, other units. Some knowledge of Java and R is desirable.
Contact: Ilze Ziedins i.ziedins@auckland.ac.nz
(Please also contact Ilze Ziedins if you are interested in applying for an HRC summer scholarship, the closing date for which is 15 September.)
Markov decision processes are widely used to determine optimal policies for stochastic systems, usually assumed to be in equilibrium. Typical examples include determining optimal admission controls for queues, and optimal routing algorithms for queueing networks. This project will investigate optimal policies for systems where demand and/or resources vary over time (for instance, arrivals varying over the course of a day). The approach will be both computational and theoretical.
Contact: Ilze Ziedins i.ziedins@auckland.ac.nz
Human genetic diversity is a fascinating topic that can be explored in a variety of different ways. Relationships in genetic data are often displayed as trees. However, the tree structure does not allow species that have diverged in genetic history to join or merge again. The tree structure does not match with modern thinking on evolution. A network, however, does not suffer from this problem. This project will involve using and expanding a large collection of DNA databases from the forensic literature with the splitstree programme (www.splitstree.org) to explore networks of human populations. This topic is ideally suited for a student who likes computing, and is willing to learn many different things.
Contact: James Curran j.curran@auckland.ac.nz
This project is offered by The Department of Statistics and Centre of Methods and Policy Application in the Social Sciences (COMPASS), Faculty of Arts, The University of Auckland.
The NZ Social Science Data Service (NZSSDS) is an ongoing initiative. The vision is that data sets, well documented with metadata for users, are made publicly available to the wider research community. Key published papers are documented along with computer code (e.g. SAS) for data manipulation and analysis. Providing all the underlying material needed to understand a piece of published research makes it more transparent, reusable and reproducible, with the potential for teaching purposes. The studentship will contribute to this system by helping to implement a template and integrate materials to the NZSSDS website.
Contact: Peter Davis pb.davis@auckland.ac.nz
This project is offered by The Department of Statistics and Centre of Methods and Policy Application in the Social Sciences (COMPASS), Faculty of Arts, The University of Auckland.
Ongoing measurement and monitoring of health system performance is crucial to improving patient outcomes and cost effectiveness. We are developing indicators (via SAS programming) from Ministry of Health administrative data (the National Minimum Data Set) and using those indicators to describe the quality of care across NZ public hospitals over the period 2001-2009. For each indicator it is hypothesised that there will be variation by patient socio-demographic characteristics and illness conditions, hospital, region and year. The studentship will involve helping to develop such indicators and examining their patterns of occurrence and distribution.
Contact: Peter Davis pb.davis@auckland.ac.nz
Texas holdem is a popular version of the card game poker.
This project is about coding a very basic version of this game in R. It will involve some calculations of probabilities (to evaluate hand values), and writing simulation and interactive graphics code in R.
Contact: Mark Holmes, m.holmes@auckland.ac.nz, Paul Murrell, p.murrell@auckland.ac.nz
One or more projects helping Chris Wild with ongoing work in the above areas. A more targeted project on investigating forms of randomisation inference that are accessible to beginners is also a possibility. All of the above is research and development work to improve statistics education at school and university levels and also to facilitate the statistics education research of Maxine Pfannkuch
Requirements: Good programming skills in R or web programming skills
Contact: Chris Wild, c.wild@auckland.ac.nz
This project involves working with Chris Wild and Alastair Scott on investigating the performance of inferential methods for sample survey and missing data problems based upon estimating equations
Requirements: good grade in STATS 310 and good programming skills in R
Contacts: Chris Wild, c.wild@auckland.ac.nz, Alastair Scott, a.scott@auckland.ac.nz
The normal Laplace distribution is a flexible distribution with four parameters which has been used in a number of contexts by Professor Bill Reed from the University of Victoria, Canada. In 2009 as a summer project, a start was made on implementing this distribution and first version of the package NormalLaplace was created. The functions implemented were just the dpqr functions and functions for the mean, variance, skewness, kurtosis and mode. Scarcely any test routines were written.
The aim of this project is to extend the functionality of the package and create suitable unit tests for all functions. Functions are required for the following:
Writing of test routines is likely to take some time.
If time permits, a start may be made on extending the package functionality to the generalized normal Laplace distribution. This is a more difficult task since no closed form exists for the density of the distribution. Numerical inversion of the characteristic function will be necessary for this case.
The package NormalLaplace is already on CRAN and a new version of the package will be submitted incorporating the results of the project.
A student who undertakes this project will gain a good knowledge of R programming and of software and tools needed for efficient software development.
Contact: David Scott, d.scott@auckland.ac.nz
Testing is vital for the creation of reliable, accurate software. One approach to test creation is the use of unit tests. In this approach for every function implemented in the software a unit test is written to test the function. When a change is made to any function, the unit tests can be run to ensure that the change made has not broken some other function. In R, the development of unit tests is facilitated by the package RUnit.
In the package GeneralizedHyperbolic only some functions have associated unit tests. This is largely historical, being due to a lack of awareness of the unit testing approach when the package was developed. The aim of this project is to complete the development of unit tests for all functions in the package.
A student who undertakes this project will gain a good knowledge of R programming and of software and tools needed for efficient software development.
Contact: David Scott, d.scott@auckland.ac.nz
Motu Kaikoura is an island off Great Barrier Island which was eradicated of rats in winter 2008. From early 2009 onwards, new rats were detected on the island. All catches were recorded and we have genetic profiles from all the early catches and many of the later catches, which gives us an unprecedented opportunity to study the genetic characteristics of a population starting from scratch. It has been suggested that the population may have started from a combination of male rats swimming from the surrounding mainland, and female rats who survived the eradication. If we can establish a sex difference among survivors and reinvaders, this has implications for island sanctuaries around the country. This involves determining where the individual rats came from, using the genetic data. The project will further aim to create a spatio-temporal genetic history of the population. We plan to use relatedness measures to establish when and where breeding began and siblings started appearing, and when and how the population developed a distinct genetic signature different from the pre-eradication signature.
Requires a student with excellent computing skills.
Contact: supervisor Rachel Fewster, r.fewster@auckland.ac.nz, co-supervisor James Russell, j.russell@auckland.ac.nz
Almost 50% of people aged over 65 eventually live in residential care facilities (rest homes or long-stay hospitals). Many are admitted from such facilities acutely to hospital, but perhaps their illness could be better managed within the facility. Who are they? Can they be characterised into one of several groupings so they could be better managed clinically? This project will involve application of regression models, hierarchical linear modelling, tree-based methods &/or other methods to analyse our existing research records. The project will be supervised by Steve Vander Hoorn and Jo Broad and will be located at North Shore Hospital, in an established health research group.
Contact: Steve Vander Hoorn, s.vanderhoorn@auckland.ac.nz
Aitken and Lucy (2004) proposed a “two level model” for the evaluation of likelihood ratios. This method uses variance components ideas along with kernel density estimation. The difficulty is making independence decisions so that the multivariate nature of the problem is tractable. I would like to investigate replacing the raw measurements with the first two (or three) principal components. This approach has an advantage in that the computations could essentially become bivariate, and that the derived variables should almost be normally distributed, which would also simplify computation. A student who is interested in this project should have done at least 302, 330 and be familiar with R.
Contact: James Curran, j.curran@auckland.ac.nz
Data has been collected from 123,968 patients attending GP practices in some areas of New Zealand (some are repeated measures within the same patient). The data has been collected with the ultimate aim of deriving cardiovascular disease risk equations in a New Zealand population, by ethnicity. The prevalence of different degrees of renal impairment is currently unknown in the New Zealand population. Both mortality (death) and morbidity (hospitalisation) information have been linked to these patients from the Ministry of Health (MOH). This project will extract all causes of renal impairment using ICD codes from the MOH to enable a variable to be coded as the degree of renal failure. In addition, serum creatinine, urinary albumin levels and estimated glomerular filtration rates are available and the relationship between these variables and the degree of renal impairment will be further investigated. These investigations will include cross-sectional associations, longitudinal associations, and Cox proportional hazards models. The statistical analyses will be carried out in the School of Population Health.
Contact: Patricia Metcalf, p.metcalf@auckland.ac.nz
Data has been collected from 123,968 patients attending GP practices in some areas of New Zealand (some are repeated measures within the same patient). The data has been collected with the ultimate aim of deriving cardiovascular disease risk equations in a New Zealand population, by ethnicity. Plasma glucose and HbA1c levels are used to diagnose diabetes. This project will carry out some exploratory analyses to describe the distribution of plasma glucose and HbA1c levels by gender, ethnicity and deprivation level and also look at the characteristics of patients with missing data versus those with non-missing data. The repeated measures nature of the data will also be further explored. The statistical analyses will be carried out in the School of Population Health.
Contact: Patricia Metcalf, p.metcalf@auckland.ac.nz
Data has been collected from 30,000 patients attending Diagnostic Medlab data collection depots during 1993 to 1994. Oral glucose tolerance tests (OGTT) are used to diagnose diabetes. More recently, the American Diabetes Association has recommended using HbA1c levels to diagnose diabetes as they are less subject to within person variation, and can be carried out in non-fasting blood sample. The relationship between the OGTT result and HbA1c levels will be further explored to determine whether an appropriate cut-off level of HbA1c can be used for diagnosing diabetes in a New Zealand population. This information has been requested by the New Zealand Society for the Study of Diabetes (NZSSD) group. The statistical analyses will be carried out in the School of Population Health.
Contact: Patricia Metcalf, p.metcalf@auckland.ac.nz
Data has been collected from 123,968 patients attending GP practices in some areas of New Zealand (some are repeated measures within the same patient). The data has been collected with the ultimate aim of deriving cardiovascular disease risk equations in a New Zealand population, by ethnicity. Serum creatinine levels and estimated glomerular filtration rates are used to determine the kidney function status of a patient. This project will carry out some exploratory analyses to describe the distribution of serum creatinine levels and estimated glomerular filtration rates by gender, ethnicity and deprivation level and also look at the characteristics of patients with missing data versus those with non-missing data. The repeated measures nature of the data will also be further explored. The statistical analyses will be carried out in the School of Population Health.
Contact: Patricia Metcalf, p.metcalf@auckland.ac.nz
In this summer school project the student would be introduced to some assessment aspects of statistics education research based on two current research projects of the supervisors.
In the research project on developing students’ risk literacy the summer school student would
In the research project on building students’ informal statistical inferential reasoning the summer school student would
Contact: Dr Stephanie Budgett, s.budgett@auckland.ac.nz; Dr Maxine Pfannkuch, m.pfannkuch@auckland.ac.nz
Whether taking selenium supplements benefits your health is still a controversial issue. A recent study in Auckland of five hundred men from Auckland tested this question. As well as pre and post supplementation blood samples many other medical and lifestyle variables were measured. The subjects were also genotyped for various polymorphisms of genes involved in selenium metabolism. The daily nutrient intake of the participants was also assessed through analysis of food diaries. The research team wish to identify which groups show significant benefits of selenium supplementation.
Supervisors: Dr Nishi Karunasinghe, Professor Lynn Ferguson, Faculty of Medicine and Health Sciences, Professor Chris Triggs
Contact: Chris Triggs, cm.triggs@auckland.ac.nz
On July 25th 2010 the WikiLeaks web site released over 90,000 US military incident reports from the war in Afghanistan. This project will explore those reports with an emphasis on data processing and data visualisation. Possible tasks include: evaluating data quality in the original files; reproducing interactive graphics like those published in the Guardian newspaper (using R); exploring new ways to visualise important features of the data. The project will involve writing R code and would suit a student with an interest in statistical computing and statistical graphics.
Contact: Paul Murrell, p.murrell@auckland.ac.nz
This project is to (i) Implement functions to compute the maximum likelihood estimates of a dozen selected univariate distributions including the Skellam, Weibull (and variants), poisson-lognormal, glog-normal, exponential-Poisson, truncated negative binomial; (ii) Investigate the use of model-selection methods such as AIC and BIC. My existing theoretical framework enables the parameters to be naturally modelled as functions of explanatory variables. Time allowing, we will also implement a multivariate technique called multidimensional scaling (MDS). The output will be additions to my VGAM package for R.
Requirements: STATS310 and STATS380 are good background, else STATS730 and STATS784. That is, a student with good grades in statistical theory and R programming.
Contact: Thomas Yee, t.yee@auckland.ac.nz
The vector generalized linear/additive (VGLM/VGAM) modelling framework as implemented by the author’s VGAM package for R fits many models by iteratively reweighted least squares. A number of topics need development and extension. These include (i) implementing the bias reduction methods by Kosmidis and Firth (2009, in Biometrika), (ii) writing software for estimating the parameters of many univariate and multivariate discrete and continuous distributions. These topics would be suitable for those with (a) a good grasp of statistical theory/inference, (b) are good at R programming. LaTeX skills would be an advantage but is unnecessary.
Requirements: STATS310 and STATS380 are good background, else STATS730 and STATS784. That is, a student with good grades in statistical theory and R programming.
Contact: Thomas Yee, t.yee@auckland.ac.nz
The 2008 World Fly Fishing Championships was held in NZ and the data was made available. Recently the 2010 World Fly Fishing Championships was also made available. This project has several facets: (i) Convert the 2010 data (in a spreadsheet) into R data frames, along with R documentation. (ii) Analyze the data further. This may mean obtaining more data from other competitions, and perform an analysis of specific competitors, possibly involving mixed models. (iii) Obtain data from Lake Otamangakau and compare the fish distributions from there relative to those caught at the WFFC. (iv) Conduct a literature search on fishing competitions in general, looking for statistical aspects. The student ideally would have had some experience in fly fishing and be competent in several branches of applied statistics, especially regression analysis.
Requirements: STATS310 and STATS380 are good background, else STATS730 and STATS784. That is, a student with good grades in statistical theory and R programming.
Contact: Thomas Yee, t.yee@auckland.ac.nz
These Summer Scholarships are available to suitable undergraduate students.
Closing date for applications: Friday 27th August 2010 (4.00 pm)
Aim: To give students research experience and acquaint them with the research activities of the Department and with potential graduate supervisors.
Emolument: $5,000 non-taxable
Duration: Full time employment for eight to twelve weeks (400 hours)
For general information regarding scholarships, please see the Scholarships and Research Grants webpage.