The group has three main (interacting) roles within Rothamsted Research:
- provision of statistical consultancy, including advice on the design of experiments and surveys, sampling and data collection, the analysis and modelling of experimental and survey data, and the interpretation and presentation of statistical analyses;
- provision of training in statistical methods for biological science researchers, including statistical summary methods, hypothesis testing, the design of experiments, the analysis of designed experiments, regression modelling, and multivariate methods);
- collaborative research, applying a wide range of statistical approaches to address scientific questions across the breadth of biological disciplines included within Rothamsted Research.
The group continues to build on a strong tradition (see 'A History of Statistics at Rothamsted') of the application of statistical approaches within research projects at Rothamsted, to make the most of the experimental resources and data collected. All research staff and students have access to one of the six consultant statisticians for advice on the design of experiments (for landscape, field, glasshouse, controlled environment and laboratory experiments - including those using high-throughput technologies), on sampling and data collection strategies, on the analysis and modelling of experimental and survey data, and on the interpretation and presentation of the results obtained from these analyses. Particular areas of statistical expertise currently include:
- the design of experiments (see 'Mead, Gilmour & Mead (2012) Statistical Principles for the Design of Experiments')
- the analysis of designed experiments (with particular application to the 'Long Term Experiments', to data from microarray gene expression studies, and to data from qRT-PCR studies)
- linear and non-linear modelling, including mixed model analysis
- modelling of non-normally distributed data (with particular application to data from the 'Rothamsted Insect Survey')
- multivariate analysis
The group has developed an annual programme of statistical training short courses, with a primary aim of providing PhD students and research staff with a sufficient understanding of key statistical ideas to enable them to apply many standard methods to their own data. The programme also aims to raise an awareness of the broader range of statistical tools that are available, to enhance the interactions that staff and students are able to have with the consultant and research statisticians within the group. The first course, Basic Statistics and Introduction to GenStat, provides a revision of key statistical concepts, using the GenStat statistical package (see 'External links' below) to manage and explore data. The second course, Design and Analysis of Simple Experiments, provides an introduction to the key concepts for the design of experiments, and explores how these concepts are applied in the analysis of data from designed experiments using analysis of variance (ANOVA). Two further courses, Introduction to Linear Regression and Advanced Regression Analysis, explore methods for the statistical modelling of biological data, including both non-linear responses and non-Normal responses. A fifth course, Introduction to Multivariate Statistics, provides guidance on this increasingly important area of statistical methodology. Responding to the changing requirements of the science community with whom we collaborate, new courses have been introduced in 2015 and 2016 - An Introduction to R provides a good grounding in this now commonly used statistical computing environment, while Geostatistics presents an introduction to this specialist area in which associated group members (Webster, Milne, Harris) have a wealth of experience.
A new textbook (Welham, Gezan, Clark & Mead (2015) Statistical Methods in Biology: Design & Analysis of Experiments and Regression), written by current and past members of the group based on materials developed for the Design and Analysis of Simple Experiments and Introduction to Linear Regression courses, has recently been published by CRC Press. The book focuses on examples and applications of statistics in the agricultural and biological sciences, presenting best practices in the design of experiments and data analysis for both designed experiments and observational studies, using the minimum required mathematical formulae, and written using language that should be familiar to agricultural and biological science researchers and post-graduate students. Associated online material includes extensive solutions to half of the exercises in the book, implementations of these solutions in three statistical pckages (GenStat, R and SAS), and guidance on using each of these packages to implement the statistical methods described in the book.
The Applied Statistics Group are also currently engaged in a BBSRC project concerned with the development of eLearning resources based on a number of the statistics training courses previously developed by the group members.
A key contribution of the group to the research activities at Rothamsted Research is through the contribution of essential statistical inputs to collaborative interdisciplinary projects. These collaborative projects may require the development of statistical theory (often involving further collaborations with academic statistical partners), or the development of software tools (usually in collaboration with VSNi, the developers of GenStat), but usually involve the modification and application of existing statistical methodology to solve novel biological problems. Current areas of interest and expertise include environmetrics, the design and analysis of microarray gene expression experiments, the (simulation) modelling of landscape scale processes associated with biodiversity and other ecosystem services, metagenomics, and analysis of RNAseq data.
Current and recent projects involving substantial inputs from group members include:
- Analysis of butterfly migratory flights using circular statistics (contact: Suzanne Clark)
- Analysis of RNAseq data (contact: Stephen Powers)
- Assessing the impact of land-use decisions on landscape-scale biodiversity (contact: Andrew Mead)
- Complex spatial variation of environmental variables: sampling, prediction and interpretation (contact: Alice Milne / Richard Webster)
- Continental-scale mapping of soil variables in Australia (contact: Richard Webster)
- Design and analysis of microarray gene expression experiments (contact: Andrew Mead)
- Design and analysis of Real-time quantitative RT-PCR (qRT-PCR) (contact: Stephen Powers)
- Emission of greenhouse gases from irrigated land (contact: Richard Webster)
- EnergyScapes and Ecosystem Services (contact: Andrew Mead)
- Exploiting yield maps and soil management zones (contact: Alice Milne)
- Greenhouse gas platform - data management of the UK national inventory (contact: Alice Milne)
- Modelling the causes of spatial and temporal variation in annual grass weeds in fields for precision weed management (contact: Helen Metcalfe / Alice Milne)
- National Capabilities: Long-term experiments and electronic Rothamsted Archive (e-RA) (contact: Rodger White)
- National Capabilities: Rothamsted Insect Survey - Annual predictions of aphid appearances (contact: Suzanne Clark)
- Reducing GHG emissions, nitrate pollution and 'lost' productivity by fully automating N fertiliser management (Auto-N) (contact: Alice Milne / Richard Webster)
- Salinity in the Yangtze delta (contact: Richard Webster)
Biomathematics and Statistics Scotland (BioSS)
British Geological Survey
University of Reading School of Agriculture, Policy and Development
University of Warwick School of Life Sciences
VSN International (VSNi)
Warwick Crop Centre
The Applied Statistics Group continues to build on the strong tradition of statistical contributions to the research at Rothamsted, which started with the appointment of Ronald A. Fisher in 1919 and continuing with many other distinguished statisticians including Frank Yates, John Nelder, John Gower and Robin Thompson. Statisticians based at Rothamsted have made major contributions to many different areas of statistical methodology, including the design and analysis of experiments, factorial experimentation, generalized linear models (GLMs), statistical modelling, multivariate analysis (particularly cluster analysis) and linear mixed models (including REML). The GenStat statistical software was mostly developed at Rothamsted, though current development is through VSN International.
Zofia Garajova (supervisor Andrew Mead, sometimes based at University of Warwick (supervisor John Clarkson))
Josh Hodge (supervisor Andrew Mead, based at University of Warwick (supervisor Rosemary Collier))
Helen Metcalfe (supervisor Alice Milne, joint with Jon Storkey (RRes - AGEC))
John Addy (supervisor Andrew Mead, joint with Mikhail Semenov (RRes - CSYS), Andy Macdonald (RRes - SSGS) and Richard Ellis (University of Reading))
Lawes Agricultural Trust (LAT) and Rothamsted Research (RR) Fellows
Roger Payne (RR)
Robin Thompson (LAT)
Richard Webster (LAT)
Darren Murray (VSNi)
Sue Welham (VSNi)
Past group members