动物栖息地选择评估的常用统计方法

利用合适的统计学方法能够更准确地理解动物的栖息地选择。本文通过对2003~2012年期间,10个国际期刊所发表的177篇关于鸟类和兽类栖息地选择论文的30种统计学方法进行分析,简要概述了目前流行的栖息地选择统计学分析方法及特点,同时对同时期的中文文献也进行了简要分析。目前关于动物栖息地选择较为流行的分析方法主要有逻辑斯蒂回归、资源选择函数、成分分析、广义线性模型、多元方差分析、基于欧几里德距离的方法、广义线性混合模型、生态位因子分析、基于个体模型、典型相关分析、物种分布模型等。广义线性模型、逻辑斯蒂回归、多元方差分析和基于欧几里德距离这些方法可以很灵活地用来分析数据,但是缺乏一个有生态学意义的理论框架。资源选择函数和生态位因子分析各自为栖息地选择研究提供了一个统一的理论框架。基于个体的模型是一个自下而上的过程,很难在系统水平形成理论。232篇国内文章中使用较多的方法是主成分分析、Mann-Whitney U检验、t检验、卡方检验、判别分析、方差分析、Vanderloeg选择系数和Scavia选择指数、逻辑斯蒂回归、Kruskal-Wallis H检验和多元回归分析等。在实际研究中,应根据所要解决的研究问题,选择切实可行的分析方法。

A Review of Statistical Methods for Evaluating Animal Habitat Selection

The use of suitable statistical models can help us to improve the understanding of species-habitat relationships. To identify the current most popular statistical methods, we surveyed the papers published in 10 SCI (Science Citation Index) journals aimed at habitat selection and journals of mainland China over the last 10 years (2003-2012). Of 30 methods used in 177 papers published in SCI journals, Logistic regression, Resource selection function, Compositional analysis, Generalized linear model, Multivariate analysis of variance, Euclidean distance-based approach, Generalized linear mixed model, Ecological-niche factor analysis, Individual-based modeling and Canonical correspondence analysis were most widely used ones. The Generalized linear model, Logistic regression, Multivariate analysis of variance and Euclidean distance-based approach were four methods very flexile when dealing with ecology data, however, the results might be lack of ecologically meanings. Resource selection function and Ecological-niche factor analysis provide us the concept which can lead to unified theory for the analysis and interpretation of habitat selection data. The Individual-based approach is a bottom-up approach which will never lead to theories at the system level. We surveyed 232 papers from mainland China, and found 19 methods were used, and the Principal component analysis, Mann-Whitney U test, Student's t test, Chi-square test, Discriminant analysis, Analysis of Variance, Vanderploeg and Scavia's first selection index, Vanderploeg and Scavia's second selection index, Logistic regression, Kruskal-Wallis H test and multiple regression analysis were most widely used.