Premium
Towards automating model selection for a mark–recapture–recovery analysis
Author(s) -
Sisson S. A.,
Fan Y.
Publication year - 2009
Publication title -
journal of the royal statistical society: series c (applied statistics)
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.205
H-Index - 72
eISSN - 1467-9876
pISSN - 0035-9254
DOI - 10.1111/j.1467-9876.2008.00656.x
Subject(s) - computer science , model selection , a priori and a posteriori , selection (genetic algorithm) , machine learning , ranking (information retrieval) , population , artificial intelligence , set (abstract data type) , mark and recapture , data mining , philosophy , demography , epistemology , sociology , programming language
Summary. Methods for fitting models to mark–recapture–recovery studies are now well established in the literature. Classical model selection methods for identifying those models which best represent the population under investigation are perhaps less satisfactory. One class of methods implements manual model searches on a model space that is restricted by strong physical understandings of the biological plausibility of each model. This can lead to highly subjective analyses requiring a priori expert knowledge, which are slow to implement and can be error prone. More automated search algorithms are now available and can be implemented with ease to consider larger classes of models. We investigate the utility of such automated algorithms and consider in particular the situation where there is a large set of near optimal models according to the model ranking function. We present a modification of an automated search procedure on an unrestricted model space and propose a procedure for model selection in the absence of a single clear optimal model. We investigate this approach through a classical mark–recapture–recovery analysis of a red deer population from the island of Rùm and conduct an investigation into senesence, which is theorized to occur in wild animal populations.