Bayesian Inference for Animal Space Use and Other Movement Metrics
Bayesian Inference for Animal Space Use and Other Movement Metrics
复制标题
动物空间使用和其他运动指标的贝叶斯推理
DOI：
--
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
C. Kuhn
中科院分区：
文献类型：
--
作者：
Devin S. Johnson;J. M. London;C. Kuhn
对动物运动和资源利用的分析已成为动物生态学研究的标准工具。遥测设备在整体尺寸和数据收集能力方面已经变得相当复杂。分析运动的统计方法已经做出反应，变得越来越复杂，通常依赖于状态空间建模。对运动指标（例如利用率分布）的估计并未效仿，主要依赖于内核密度估计。在这里，我们考虑一种用于推断空间使用的方法，该方法不存在与利用率分布的核密度估计相关的所有主要问题，例如自相关、不规则时间间隙和观测位置的误差。我们提出的方法基于数据增强方法，该方法将使用定义为仅部分观察到的动物完整路径的摘要。我们使用完整路径的后验分布中的样本来构建感兴趣度量的后验样本。提出并比较了三种基于重要性采样的基本方法，用于从路径的后验分布中进行采样。我们通过估计阿拉斯加普里比洛夫群岛雌性北方海狗的潜水强度空间图来演示增强方法。
The analysis of animal movement and resource use has become a standard tool in the study of animal ecology. Telemetry devices have become quite sophisticated in terms of overall size and data collecting capacity. Statistical methods to analyze movement have responded, becoming ever more complex, often relying on state-space modeling. Estimation of movement metrics such as utilization distributions have not followed suit, relying primarily on kernel density estimation. Here we consider a method for making inference about space use that is free of all of the major problems associated with kernel density estimation of utilization distributions such as autocorrelation, irregular time gaps, and error in observed locations. Our proposed method is based on a data augmentation approach that defines use as a summary of the complete path of the animal which is only partially observed. We use a sample from the posterior distribution of the complete path to construct a posterior sample for the metric of interest. Three basic importance sampling based methods for sampling from the posterior distribution of the path are proposed and compared. We demonstrate the augmentation approach by estimating a spatial map of diving intensity for female northern fur seals in the Pribilof Islands, Alaska.