Two case studies detailing Bayesian parameter inference for dynamic energy budget models
Two case studies detailing Bayesian parameter inference for dynamic energy budget models
复制标题
DOI：
10.1016/j.seares.2018.07.014
复制
发表时间：
2019-01-01
影响因子：
2
通讯作者：
Johnson, Leah R.
中科院分区：
文献类型：
--
作者：
Boersch-Supan, Philipp H.;Johnson, Leah R.
个体生活史轨迹的机械表示是预测生物体在新环境条件下生长、繁殖和生存的有力工具。动态能量收支（DEB）理论提供了紧凑的模型来描述生物体在其整个生命周期中能量的获取和分配。然而，估计DEB模型参数及其相关的不确定性和协方差并不是微不足道的。贝叶斯推理提供了一种一致的方法来估计参数的不确定性，并通过模型传播它，同时还利用先验信息来约束参数空间。我们概述了能源预算模型的贝叶斯推理方法，并提供了两个案例研究-基于一个简化的DEBkiss模型，和标准的DEB模型-详细介绍了使用开源软件包deBInfer的推理过程的实施。我们演示了如何在贝叶斯框架中估计DEB和DEBkiss参数，但我们的研究结果也突出了识别DEB模型参数的困难，这提醒我们，拟合这些模型需要统计谨慎。
Mechanistic representations of individual life-history trajectories are powerful tools for the prediction of organismal growth, reproduction and survival under novel environmental conditions. Dynamic energy budget (DEB) theory provides compact models to describe the acquisition and allocation of energy by organisms over their full life cycle. However, estimating DEB model parameters, and their associated uncertainties and covariances, is not trivial. Bayesian inference provides a coherent way to estimate parameter uncertainty, and propagate it through the model, while also making use of prior information to constrain the parameter space. We outline a Bayesian inference approach for energy budget models and provide two case studies - based on a simplified DEBkiss model, and the standard DEB model - detailing the implementation of such inference procedures using the open-source software package deBInfer. We demonstrate how DEB and DEBkiss parameters can be estimated in a Bayesian framework, but our results also highlight the difficulty of identifying DEB model parameters which serves as a reminder that fitting these models requires statistical caution.