A critical comparison of integral projection and matrix projection models for demographic analysis: Comment
A critical comparison of integral projection and matrix projection models for demographic analysis: Comment
复制标题
用于人口分析的积分投影和矩阵投影模型的关键比较：评论
DOI：
10.1002/ecy.3605
复制
发表时间：
2022
期刊：
影响因子：
4.8
通讯作者：
Rees, Mark
中科院分区：
文献类型：
--
作者：
Ellner, Stephen P.;Adler, Peter B.;Childs, Dylan Z.;Hooker, Giles;Miller, Tom E.;Rees, Mark
Doak et al.（2021，以下简称D21）中最令人不安的一句话是：“最后，我们中的一些人（Doak, Morris， Garcia等人）……我们发现，审稿人和编辑越来越倾向于将这两种方法区分开来，认为矩阵模型已经过时了。”我们完全同意不应该有“对IPM的崇拜”。种群生态学家使用矩阵预测模型（MPMs）、积分预测模型（IPMs）、微分方程、阶段结构和生理结构模型、基于主体的模型等等。都有自己的位置。如果您的模型对于您的物种、数据和问题来说是一个糟糕的选择，或者它没有很好地参数化，这是一个合理的批评。“过时”却不是。然而，正如D21所断言的那样，我们并不知道有任何论文实际上“质疑基于矩阵模型的许多研究的有用性”，声称ipm普遍优于mpm，或者断言需要“基本的重新工具”。最近许多使用COMPADRE/COMADRE矩阵模型数据库的比较研究（Salguero-Gómez等人，2015年，2016年），包括Adler等人（2014年）的工作表明，很少有生态学家真正质疑mpm的有用性。因此，在已发表的文献中，我们没有看到继续使用mpm的真正威胁。当然，编辑和审稿人如何解读文学是另一回事。我们进一步同意D21的观点，即mpm和ipm不应该被二分类，而是对任何结构化群体进行建模应该涉及一组共同的考虑因素和经验法则。尽管提倡一种更统一的建模框架，D21仍然认为在他们的图1中，这些是截然不同的竞争方法，建模工作流在MPM和IPM之间的初始选择上分叉。作为对比，我们建议一个强调从数据收集到模型分析的单一考虑流的工作流（图1）。根据物种、数据集和问题，我们的工作流程可以产生IPM、MPM或另一个结构化的种群模型。我们使用这个图表（图1）来构建两个中心建议，它们代表了我们与D21给出的指导意见的主要分歧。
The most troubling sentence in Doak et al.(2021, hereafter D21) is this:“Finally, some of us (Doak, Morris, Garcia pers. obs.) have seen an increasing tendency of reviewers and editors to dichotomize these two approaches, considering matrix models as out of fashion.” We fully agree that there should not be a “cult of IPM”. Population ecologists use matrix projection models (MPMs), integral projection models (IPMs), differential equations, stage-structured and physiologically structured models, agent-based models, and more. All have a place. If your model is a poor choice for your species, data, and questions, or it has been poorly parameterized, that is a legitimate criticism.“Out of fashion” is not. However, we are not aware of any papers actually “calling into question the usefulness of the many studies based on matrix models” as D21 assert, claiming that IPMs are universally superior to MPMs, or asserting that a “fundamental re-tooling” is needed. Numerous recent comparative studies using the COMPADRE/COMADRE matrix model databases (Salguero-Gómez et al., 2015, 2016), including work by Adler et al (2014), suggest that few ecologists actually question the usefulness of MPMs. So, in the published literature, we see no real threat to the continued use ofMPMs. How editors and reviewers interpret that literature is, of course, a separate matter. We further agree with D21 that MPMs and IPMs should not be dichotomized, rather modeling any structured population should involve a common set of considerations and rules of thumb. Despite advocating a more unified modeling framework, D21 perpetuate the idea that these are distinct competing approaches in their figure 1, a modeling workflow that forks at an initial choice between MPM and IPM. As a counterpoint, we suggest a workflow emphasizing a single stream of considerations from data collection through model analysis (Figure 1). Depending on the species, data set, and question, our workflow could produce an IPM, MPM, or another structured population model. We use this diagram (Figure 1) to frame two central recommendations that represent our primary disagreements with the guidance given by D21.