Conditional likelihood approach for analyzing single visit abundance survey data in the presence of zero inflation and detection error
Conditional likelihood approach for analyzing single visit abundance survey data in the presence of zero inflation and detection error
复制标题
在零膨胀和检测误差存在的情况下分析单次访问丰度调查数据的条件似然方法
DOI：
--
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
E. Bayne
中科院分区：
文献类型：
--
作者：
P. Sólymos;S. Lele;E. Bayne
当前校正检测误差的方法需要多次访问同一勘测位置。存在许多仅使用单次访视收集的历史数据集，并且后勤/成本考虑阻止许多当前的研究计划收集多次访视数据。在本文中，我们探讨可以做什么单次访问计数数据时，有检测错误。我们表明，当影响检测和丰度的适当协变量可用时，条件似然可用于估计二项零膨胀泊松（ZIP）混合模型的回归参数并校正检测误差。我们使用观察到的Ovenbirds（Seiurus aurocapilla）的数量来说明二项式零膨胀泊松混合模型的参数估计，该模型使用来自最大和最长的生态时间序列数据集之一的数据子集，该数据集只有一次访问。我们的单次访问方法具有以下特点：（i）它不需要假设一个封闭的人口或调整所造成的运动或迁移;（ii）它是成本效益，使生态学家覆盖更大的地理区域比可能的时候，必须返回网站;和（iii）由此产生的估计似乎是统计和计算效率很高。版权所有© 2012约翰威利父子有限公司.
Current methods to correct for detection error require multiple visits to the same survey location. Many historical datasets exist that were collected using only a single visit, and logistical/cost considerations prevent many current research programs from collecting multiple visit data. In this paper, we explore what can be done with single visit count data when there is detection error. We show that when appropriate covariates that affect both detection and abundance are available, conditional likelihood can be used to estimate the regression parameters of a binomial–zero‐inflated Poisson (ZIP) mixture model and correct for detection error. We use observed counts of Ovenbirds (Seiurus aurocapilla) to illustrate the estimation of the parameters for the binomial–zero‐inflated Poisson mixture model using a subset of data from one of the largest and longest ecological time series datasets that only has single visits. Our single visit method has the following characteristics: (i) it does not require the assumptions of a closed population or adjustments caused by movement or migration; (ii) it is cost effective, enabling ecologists to cover a larger geographical region than possible when having to return to sites; and (iii) its resultant estimators appear to be statistically and computationally highly efficient. Copyright © 2012 John Wiley & Sons, Ltd.