Author: Kristylee Marr and Yvonne Buckley
Honours student Kristylee Marr and Yvonne attended the EcoStats Symposium at the University of New South Wales on 11-12 July 2013 (http://www.eco-stats.unsw.edu.au/symposium.html) . This event was organised by ecologist and statistician David Warton (http://www.eco-stats.unsw.edu.au/) discuss current issues and future directions at the ecology-statistics interface.
There was a strong species distribution modelling flavour to the first day. There is a huge demand from ecologists for methods for building niche/distribution models using widely available but problematic presence-only data. This has led to philosophical differences in the field, alternative methodologies and numerous commentaries and as one speaker put it “much heat and little light”.
Newcomers to this field like honours student, Kristylee, quickly find out that there is a lot of uncertainty involved in species distribution modelling, and that the inner workings of specialist models like MaxEnt – a method that builds models based on maximum entropy – can be obscure. Ecologists may not understand as much about MaxEnt as they do about more general and familiar models such as GLMs. Speakers at the symposium highlighted the similarity between Maximum Entropy models and Inhomogeneous Poisson Process (IPP) models (a type of Generalised Linear Model (GLM)) for presence only data. Happy days! Much is known about GLMs and this result gives us clarity and a body of theory to support our model constructions. There is the potential now to dive in further and explore the possibilities within GLMs and their diagnostic tools to extend our ability to model presence data more effectively.