FN Archimer Export Format
PT J
TI Variable Selection and Accurate Predictions in Habitat Modelling: a Shrinkage Approach
BT 
AF AUTHIER, Matthieu
   SARAUX, Claire
   PERON, Clara
AS 1:1,5;2:2;3:3,4,5;
FF 1:;2:PDG-RBE-MARBEC-LHM;3:;
C1 Univ La Rochelle, Observ PELAGIS UMS 3462, La Rochelle, France.
   Ifremer Inst Francais Rech Exploitat Mer, UMR MARBEC, Sete, France.
   Univ Tasmania, Inst Marine & Antarctic Studies, Kingston, Tas, Australia.
   Australian Antarctic Div, Kingston, Tas, Australia.
   Ctr Ecol Fonct & Evolut, Ecol Spatiale Populat, Montpellier, France.
C2 UNIV LA ROCHELLE, FRANCE
   IFREMER, FRANCE
   UNIV TASMANIA, AUSTRALIA
   AUSTRALIAN ANTARCTIC DIV, AUSTRALIA
   CEFE, FRANCE
SI SETE
SE PDG-RBE-MARBEC-LHM
UM MARBEC
IN WOS Ifremer jusqu'en 2018
   copubli-france
   copubli-univ-france
   copubli-int-hors-europe
IF 4.52
TC 13
UR https://archimer.ifremer.fr/doc/00335/44590/44307.pdf
   https://archimer.ifremer.fr/doc/00335/44590/107910.pdf
LA English
DT Article
CR PELMED 2011
BO L'Europe
AB Habitat modelling is increasingly relevant in biodiversity and conservation studies. A typical application is to predict potential zones of specific conservation interest. With many environmental covariates, a large number of models can be investigated but multi-model inference may become impractical. Shrinkage regression overcomes this issue by dealing with the identification and accurate estimation of effect size for prediction. In a Bayesian framework we investigated the use of a shrinkage prior, the Horseshoe, for variable selection in spatial generalized linear models (GLM). As study cases, we considered 5 datasets on small pelagic fish abundance in the Gulf of Lion (Mediterranean Sea, France) and 9 environmental inputs. We compared the predictive performances of a simple kriging model, a full spatial GLM model with independent normal priors for regression coefficients, a full spatial GLM model with a Horseshoe prior for regression coefficients and 2 zero-inflated models (spatial and non-spatial) with a Horseshoe prior. Predictive performances were evaluated by cross-validation on a hold-out subset of the data: models with a Horseshoe prior performed best, and the full model with independent normal priors worst. With an increasing number of inputs, extrapolation quickly became pervasive as we tried to predict from novel combinations of covariate values. By shrinking regression coefficients with a Horseshoe prior, only one model needed to be fitted to the data in order to obtain reasonable and accurate predictions, including extrapolations.
PY 2017
PD APR
SO Ecography
SN 0906-7590
PU Wiley
VL 40
IS 4
UT 000400176500010
BP 549
EP 560
DI 10.1111/ecog.01633
ID 44590
ER

EF