ObjectivesA thorough initial training for intending professional statisticians and a solid foundation for further study or research in applied or theoretical statistics.
Entry requirementsEntry requirements A first or second class Honours degree with substantial mathematical and statistical components.
Academic titleMSc Statistics
Course descriptionMSc in Statistics
The MSc in Statistics provides both a practically based professional statistical training and a foundation for those wishing to pursue further research. It is available via distance learning (2 or 3 years, part-time) as well as residential study (1 year full-time).
The MSc is a well-established course, and its quality has long been recognised by external bodies, including EPSRC. Bursaries are now available for distance learning, as well as studentships and bursaries for residential study.
Modules
- Data Analysis. This module casts participants in the role of practising statisticians who are faced with projects from a variety of fields. Each project will be written up in a report and some involve other elements (for example, oral presentations and role play). Most projects are based on actual problems and data, often originating from consulting activities, and are 'open ended' with no obviously appropriate method. The student is required to exercise general statistical and scientific skills and some ability in computing to produce suitable written and oral reports. A diversity of approaches to data analysis is encouraged, as is real insight into the problem; mere technical manipulation is discouraged. Several projects require collaborative work by students, working in teams of three or four, and one involves the team presenting their results to the 'clients' in a role-playing exercise. This last provides an excellent opportunity to encounter issues associated with costing work and statistical ethics.
- Statistical Laboratory. This module provides an overview of many statistical techniques, with an emphasis on practical implementation and interpretation of results rather than theoretical justifications. S-PLUS is the main computing tool, with a major aim being to produce competent users. However SAS is important in some application areas and so students will also attain a level of competence in that package. Areas covered include: Exploratory Data Analysis, non-parametrics (including density estimation), generalized linear modelling, survival data, time series, multivariate topics, tree-based methods, robust methods, local regression, spatial data, non-linear regression and generalized additive modelling. The module occupies the middle ground between the open-ended project module (PAS6001 Data Analysis) and the more theoretical development in modules on specific topics. Its role is to enhance the student's practical statistical competence and to develop confidence in acquiring new techniques.
- Linear Modelling. The module covers linear and generalized linear models, stressing the power of linear modelling structures in their many guises. Topics include: the general linear model, diagnostics, model building, generalised linear models, binary data and logistic regression, count data, log-linear models and contingency tables, random effects and multi-level modelling. This module provides a wide ranging coverage of central topics for any applied statistician.
- Inference. The keynote of this module is the principles and the methods of implementation of both frequentist and Bayesian inference. Topics in frequentist inference include likelihood, sufficiency and ancillarity, construction of point and interval estimators, likelihood tests, profile likelihood, and asymptotics. The Bayesian inference section considers prior distributions and prior modelling, principles of elicitation, posterior summarization, predictive inference and decisions. Computational techniques are emphasized throughout and the unit includes discussion of general numerical methods, the Expectation Maximization algorithm, simulation and Markov Chain Monte Carlo.
- Dependent Data. This module develops the ideas needed to deal with data sets with complex structure that are a feature of most real applications. It begins with a practical introduction to multivariate analysis: graphical methods, EDA and dimension reduction, principal components, MANOVA and discrimination. Another major component is time series: models, model fitting and forcasting, including both Box-Jenkins and state space models. Topics extending these basic ideas may include hierarchical models and repeated measures
- Sampling, Design, Medical Statistics. This hybrid module examines the issues involved in collecting data and gives a flavour of one of the important applications areas of the subject. The Sampling Theory section covers simple random sampling, stratification, sample sizes and practical issues in sample surveys. The Design portion looks at blocking, randomization, factorials, designing for real problems and the theory of block designs. The Medical section considers first issues in clinical trials such as: design of clinical trials, controls, protocols, blind and double blind arrangements, ethics, protocol deviations, sample size, cross-over trials, and meta-analysis; and then looks at the special features of the analysis of survival data