Is everything a hidden process model? Case studies in ecology
Ruth King, University of St Andrews
Venue: Room A54, Postgraduate Statistics Centre, Lancaster University
Date: 19-06-2014, 4 - 5pm
Numerous statistical models have been developed within the area of statistical ecology to fit to a wide range of experimental data. The corresponding associated likelihood functions have been derived and inference typically obtained either via maximum likelihood, or more recently, using Bayesian methods. Many of the models and model-fitting techniques have been developed independently of each other. However, increasingly, many of these models (traditionally regarded as different) can be viewed as simple hidden process models – and most notably as a hidden Markov model. This leads to a simple unification of many of the different statistical models under a common umbrella and permits standard model-fitting tools and techniques to be applied.
I will briefly describe the general hidden process (i.e. Markov) model structure before describing how a range of standard ecological models fit into such a framework. Finally, I will conclude with some new (cool!) tools that can be applied to the models using advanced (albeit standard) hidden Markov model techniques addressing long-standing problems within statistical ecology. I will demonstrate these advanced techniques using real examples.