Why environmental scientists are becoming Bayesians
Why environmental scientists are becoming Bayesians
复制标题
DOI：
10.1111/j.1461-0248.2004.00702.x
复制
发表时间：
2005-01-01
期刊：
影响因子：
8.8
通讯作者：
Clark, JS
中科院分区：
文献类型：
--
作者：
Clark, JS
计算统计学的进步为生态推断和预测通常所需的高维模型提供了一个通用框架。分层贝叶斯（HB）表示一种建模结构，能够利用不同的信息源，适应未知（或不可知）的影响，并对描述复杂关系的大量潜在变量和参数进行推断。在这里，我总结了HB的结构，并提供了常见的时空问题的例子。灵活的框架意味着参数、变量和潜变量可以代表比传统模型中处理的更广泛的模型元素类别。推断和预测依赖于两种类型的随机性，包括（1）不确定性，它描述了我们对固定量的知识，它适用于所有“不可观测”（潜变量和参数），并且它随着样本量的增加而逐渐下降，以及（2）可变性，适用于无法用确定性过程解释的波动，并且不会随着样本量的增加而逐渐下降。例子表明如何不同的来源的随机性影响的推断和预测，以及如何津贴随机影响可以指导研究。
Advances in computational statistics provide a general framework for the high-dimensional models typically needed for ecological inference and prediction. Hierarchical Bayes (HB) represents a modelling structure with capacity to exploit diverse sources of information, to accommodate influences that are unknown (or unknowable), and to draw inference on large numbers of latent variables and parameters that describe complex relationships. Here I summarize the structure of HB and provide examples for common spatiotemporal problems. The flexible framework means that parameters, variables and latent variables can represent broader classes of model elements than are treated in traditional models. Inference and prediction depend on two types of stochasticity, including (1) uncertainty, which describes our knowledge of fixed quantities, it applies to all 'unobservables' (latent variables and parameters), and it declines asymptotically with sample size, and (2) variability, which applies to fluctuations that are not explained by deterministic processes and does not decline asymptotically with sample size. Examples demonstrate how different sources of stochasticity impact inference and prediction and how allowance for stochastic influences can guide research.