Glen D 16 Nov 2009

Monday 16 November 2009, 11:30 AM

Glen De'ath

Australian Institute of Marine Science

Modern statistical methods to explain and predict complex ecosystems

Understanding our ecology and environment is essential to our survival. Such understanding must include the capacity to both explain and predict our marine and terrestrial ecosystems. Modern statistical models offer such capacity. In this talk I will outline the historical development of statistics from Sir Ronald Fisher to machine learning, and how the objectives of statistical analysis have changed from an emphasis on hypothesis tests to that of model selection, explanation and prediction. A family of methods based on classification and regression trees have been developed since the mid 1980s. These include bagging and boosting of trees and will be outlined in a convivial non-technical manner and contrasted to more traditional methods. We will see how these new statistical methods address the issues of explanation and prediction. Finally, a complex data set on coral reef biodiversity will be used to illustrate the application of methods.

Bio – Interests: Statistics, Mathematics, Computation, History and Philosophy of Science. Degrees: BSc Mathematics; University of Newcastle, UK. MSc Fluid mechanics and Oceanography; Uni of NSW Sydney. MEd Social Psychology; Uni of WA, Perth Australia Australia. PhD Ecological Statistics; James Cook Uni, Townsville, Australia. Employment: Universities in Australia from 1973-2000; Universities of New South Wales, Western Australia and South Australia, and James Cook University, North Queensland. Australian Institute of Marine Science from 2000.