Following a recent successful bid for funding to Natural Environment Research Council, the following training course and workshops are being offered in the University of Glasgow. Further details and registration can be found at http://www.gla.ac.uk/schools/mathematicsstatistics/events/conferences/
Core and advanced statistical training courses for environmental scientists
11-12 December 2013 Flexible regression modelling
This workshop will introduce some of the theory and application of advanced regression models including non-linear, nonparametric and generalised additive models in environmental contexts. A variety of approaches for smoothing including local polynomial regression and regression splines will be explained and illustrated through a series of lectures and practical lab sessions. The workshop will illustrate the appropriate uses and restrictions of advanced regression models, using R. Participants will develop appropriate methods for the construction, selection and evaluation of advanced regression models, and will meet the application of advanced regression models in a variety of practical environmental contexts. The workshop will also discuss the problems created by data which are dependent, missing and above/below the limit of detection.
6-10 January 2014 Quantifying the environment
This residential 1 week long training course takes participants potentially new to statistics and provides core training in theory, modelling and computation. It builds on our experience of delivering such a course for NERC supported students over the past 10 years, the feedback for which has been universally excellent. The course will cover a range of topics starting with basic statistical inference (estimation, confidence intervals, etc) and including more advanced topics such as trend analysis and time series, modern adaptive regression, Bayesian methods. Throughout the course students will be immersed in modern statistical computation using R, which will provide transferrable skills in scientific computing. The overarching goal is to offer practical training, so that as well as covering specific skills, the students will also be trained in the actual implementation and interpretation of the analysis. The training course has a number of component parts: expository lectures, case studies, practical computer-based sessions and informal discussion sessions (including review of statistical analyses in the environmental literature) with the emphasis on problem solving.
3-4 February 2014 Spatio-temporal modelling
This workshop will introduce statistical approaches to modelling data that have spatial and temporal structure. The workshop will first give indepth discussions of purely spatial and purely temporal modelling, including : geostatistics (including Kriging), areal (lattice) models including Markov Random fields and point process models including homogeneous and inhomogeneous Poisson processes; and: autoregressive moving average models; trend detection and estimation and prediction. These will then be extended to introduce spatio-temporal modelling, dealing first with separable spatial and temporal correlation structures, before finally addressing a full spatio-temporal construction.
March 2014 Functional data analysis
This workshop will introduce methods in functional data analysis, with an emphasis on practical issues and data arising from environmental monitoring devices and optical or mechanical tracking devices. Functional data analysis is a new and very powerful statistical methodology, which treats time series data in new ways (the “datapoint” becomes the curve).
The workshop will train students to identify scenarios where data may be considered to be smooth functions and construct visualization strategies and implement nonparametric smoothing for exploring functional data. Using several environmental data sets we will illustrate ways to describe the variation among a group of curves, to describe differences between groups of curves and to understand the effect of one set of curves on another by formulating and fitting several types of functional linear models. We also discuss some techniques that are unique to functional data: curve alignment and the analysis of rates of changes or derivatives.