Evaluating resource selection functions

被引:1909
作者
Boyce, MS [1 ]
Vernier, PR
Nielsen, SE
Schmiegelow, FKA
机构
[1] Univ Alberta, Dept Sci Biol, Edmonton, AB T6G 2E9, Canada
[2] Univ British Columbia, Dept Forest Sci, Vancouver, BC V6T 1Z4, Canada
[3] Univ Alberta, Dept Renewable Resources, Edmonton, AB T6G 2H1, Canada
基金
美国国家科学基金会; 加拿大自然科学与工程研究理事会;
关键词
habitat selection; logistic regression; model selection; prediction; RSF; resource selection functions; validation;
D O I
10.1016/S0304-3800(02)00200-4
中图分类号
Q14 [生态学(生物生态学)];
学科分类号
071012 ; 0713 ;
摘要
A resource selection function (RSF) is any model that yields values proportional to the probability of use of a resource unit. RSF models often are fitted using generalized linear models (GLMs) although a variety of statistical models might be used. Information criteria such as the Akaike Information Criteria (AIC) or Bayesian Information Criteria (BIC) are tools that can be useful for selecting a model from a set of biologically plausible candidates. Statistical inference procedures, such as the likelihood-ratio test, can be used to assess whether models deviate from random null models. But for most applications of RSF models, usefulness is evaluated by how well the model predicts the location of organisms on a landscape. Predictions from RSF models constructed using presence/absence (used/ unused) data can be evaluated using procedures developed for logistic regression, such as confusion matrices, Kappa statistics, and Receiver Operating Characteristic (ROC) curves. However, RSF models estimated from presence/ available data create unique problems for evaluating model predictions. For presence/available models we propose a form of k-fold cross validation for evaluating prediction success. This involves calculating the correlation between RSF ranks and area-adjusted frequencies for a withheld sub-sample of data. A similar approach can be applied to evaluate predictive success for out-of-sample data. Not all RSF models are robust for application in different times or different places due to ecological and behavioral variation of the target organisms. (C) 2002 Elsevier Science B.V. All rights reserved.
引用
收藏
页码:281 / 300
页数:20
相关论文
共 67 条
[1]   DANGERS OF USING OPTIMAL CUTPOINTS IN THE EVALUATION OF PROGNOSTIC FACTORS [J].
ALTMAN, DG ;
LAUSEN, B ;
SAUERBREI, W ;
SCHUMACHER, M .
JOURNAL OF THE NATIONAL CANCER INSTITUTE, 1994, 86 (11) :829-835
[2]   Null hypothesis testing: Problems, prevalence, and an alternative [J].
Anderson, DR ;
Burnham, KP ;
Thompson, WL .
JOURNAL OF WILDLIFE MANAGEMENT, 2000, 64 (04) :912-923
[3]  
[Anonymous], [No title captured]
[4]  
[Anonymous], 2001, STAT STAT SOFTW REL
[5]   DETERMINING SPECIES RESPONSE FUNCTIONS TO AN ENVIRONMENTAL GRADIENT BY MEANS OF A BETA-FUNCTION [J].
AUSTIN, MP ;
NICHOLLS, AO ;
DOHERTY, MD ;
MEYERS, JA .
JOURNAL OF VEGETATION SCIENCE, 1994, 5 (02) :215-228
[6]   MEASUREMENT OF THE REALIZED QUALITATIVE NICHE - ENVIRONMENTAL NICHES OF 5 EUCALYPTUS SPECIES [J].
AUSTIN, MP ;
NICHOLLS, AO ;
MARGULES, CR .
ECOLOGICAL MONOGRAPHS, 1990, 60 (02) :161-177
[7]  
Boyce M. S., 2000, GRIZZLY BEAR RECOVER, P231
[8]   Relating populations to habitats using resource selection functions [J].
Boyce, MS ;
McDonald, LL .
TRENDS IN ECOLOGY & EVOLUTION, 1999, 14 (07) :268-272
[9]   Relating populations to habitats - Reply [J].
Boyce, MS ;
McDonald, LL ;
Manly, BFJ .
TRENDS IN ECOLOGY & EVOLUTION, 1999, 14 (12) :490-490
[10]  
BOYCE MS, 2001, J AGR BIOL ENVIR ST, V7, P51