Of Uberfleas and Krakens: Detecting Trade-offs Using Mixed Models
Of Uberfleas and Krakens: Detecting Trade-offs Using Mixed Models
复制标题
DOI：
10.1093/icb/icx015
复制
发表时间：
2017-08-01
影响因子：
2.6
通讯作者：
Wilson, Robbie S.
中科院分区：
文献类型：
--
作者：
Careau, Vincent;Wilson, Robbie S.
所有动物在完成诸如捕获猎物、保卫领地、获得配偶和逃避捕食者等任务时都会经历性能权衡。那么，为什么在整个生物体水平上检测性能权衡是如此困难？为什么我们有时甚至会在两个被预测为负相关的绩效特质之间获得正相关？在这里，我们探索两个合理的解释。首先，大多数分析是基于个体最大值（即，个人最佳），这可能在相关性估计中引入偏差。第二，当对比过程发生在个体内与个体内水平时，表型相关性本身可能是权衡的不良指标。其中一种情况是在生命史理论中发展的“大房子大汽车”模型，该模型解释了“超级跳蚤”的存在，这些跳蚤在各个方面都是上级的（因为它们比其他跳蚤获得更多的资源）。我们强调，完全相反的情况下可能会发生性能权衡，在个人的权衡可能会被掩盖在个人的身体状况的变化。在这些替代方案中进行测试的最佳方法之一是收集重复的性能特征对，并使用多变量混合模型（MMM）进行分析。MMM允许直接和同时检查的性状相关性在个体和个体内的水平。我们使用一个简单的模拟工具（R中的SQuID包）来创建一个Krakens种群，Krakens是一种神话中的巨型乌贼状海洋生物，其形态在游泳速度和力量或击沉船只的能力之间产生性能权衡。模拟表明，使用个体最大值会引入偏差，当个体的重复样本数量（n（试验））不同时，这种偏差尤其严重。最后，我们展示了MMM如何帮助检测性能（或任何其他类型的）权衡，并提供额外的见解（例如，帮助检测可塑性整合）。我们希望研究人员在探索整个动物表现的权衡时采用MMM。
All animals experience performance trade-offs as they complete tasks such as capturing prey, defending territories, acquiring mates, and escaping predators. Why then, is it so hard to detect performance trade-offs at the whole-organismal level? Why do we sometimes even obtain positive correlations between two performance traits that are predicted to be negatively associated? Here we explore two plausible explanations. First, most analyses are based on individual maximal values (i.e., personal best), which could introduce a bias in the correlation estimates. Second, phenotypic correlations alone may be poor indicators of a trade-off when contrasting processes occur at the among-versus within-individual levels. One such scenario is the "big houses big cars" model developed in life-history theory to explain the existence of "uberfleas" that are superior in all regards (because they acquire more resources than others). We highlight that the exact opposite scenario might occur for performance trade-offs, where among-individual trade-offs may be masked by within-individual changes in physical condition. One of the best ways to test among these alternative scenarios is to collect repeated pairs of performance traits and analyze them using multivariate mixed models (MMMs). MMMs allow straightforward and simultaneous examination of trait correlations at the among-and within-individual levels. We use a simple simulation tool (SQuID package in R) to create a population of Krakens, a mythical giant squid-like sea creature whose morphology generates a performance trade-off between swimming speed and strength or ability to sink ships. The simulations showed that using individual maximum values introduces a bias that is particularly severe when individuals differ in the number of repeated samples (n(trial)). Finally, we show how MMMs can help detect performance (or any other type of) trade-offs and offer additional insights (e.g., help detect plasticity integration). We hope researchers will adopt MMMs when exploring trade-offs in whole-animal performances.