Social network dynamics: the importance of distinguishing between heterogeneous and homogeneous changes.
Social network dynamics: the importance of distinguishing between heterogeneous and homogeneous changes.
复制标题
DOI：
10.1007/s00265-015-2030-x
复制
发表时间：
2015-12
影响因子：
2.3
通讯作者：
Alberts SC
中科院分区：
文献类型：
--
作者：
Franz M;Alberts SC
社会网络分析越来越多地被应用于理解动物社会性的进化。确定复杂社会结构的生态和进化驱动因素需要推断社会网络如何随着时间的推移而变化。在大多数观察性研究中，抽样误差可能会影响网络的表观结构。在此，我们认为，当社会网络随时间变化时，现有方法往往不能对某些类型的抽样误差进行足够的控制。具体地说，我们认为社交网络中可能会发生两种不同类型的变化，异质变化和同质变化，理解网络动态需要区分这两种不同类型的变化，这两种变化并不是相互排斥的。如果关系发生了不同的变化，例如，如果一些关系终止了，但另一些关系保持不变，就会发生异质变化。如果所有的关系都以同样的方式成比例地受到影响，例如，如果所有二人的美容率都下降了，那么就会发生同质的变化。关系强度的同质性下降可以极大地降低观察到弱关系的概率，从而产生异质网络变化的外观。通过模拟，我们确认如果不能区分同质和异质变化，可能会导致关于网络动力学的错误结论。我们还表明，Bootstrap测试无法区分同构和异质更改。作为这一问题的解决方案，我们表明，适当的随机化检验可以推断是否发生了异质性变化。最后，我们使用一个野生狒狒的经验数据集进行了一个例子分析，说明了使用随机化检验的实用性。
Social network analysis is increasingly applied to understand the evolution of animal sociality. Identifying ecological and evolutionary drivers of complex social structures requires inferring how social networks change over time. In most observational studies, sampling errors may affect the apparent network structures.Here, we argue that existing approaches tend not to control sufficiently for some types of sampling errors when social networks change over time. Specifically, we argue that two different types of changes may occur in social networks, heterogeneous and homogeneous changes, and that understanding network dynamics requires distinguishing between these two different types of changes, which are not mutually exclusive. Heterogeneous changes occur if relationships change differentially, e.g. if some relationships are terminated but others remain intact. Homogeneous changes occur if all relationships are proportionally affected in the same way, e.g. if grooming rates decline similarly across all dyads. Homogeneous declines in the strength of relationships can strongly reduce the probability of observing weak relationships, producing the appearance of heterogeneous network changes. Using simulations, we confirm that failing to differentiate homogeneous and heterogeneous changes can potentially lead to false conclusions about network dynamics. We also show that bootstrap tests fail to distinguish between homogeneous and heterogeneous changes. As a solution to this problem we show that an appropriate randomization test can infer whether heterogeneous changes occurred. Finally, we illustrate the utility of using the randomization test by performing an example analysis using an empirical data set on wild baboons.
登录
查看更多内容
影响因子：
2.2
作者：
Franz, Mathias;Altmann, Jeanne;Alberts, Susan C.
通讯作者：
Alberts, Susan C.
影响因子：
3.1
作者：
Snijders, Tom A. B.;van de Bunt, Gerhard G.;Steglich, Christian E. G.
通讯作者：
Steglich, Christian E. G.
影响因子：
2.5
作者：
Ansmann, Ina C.;Parra, Guido J.;Lanyon, Janet M.
通讯作者：
Lanyon, Janet M.
DOI：
10.1098/rspb.2013.1113
发表时间：
2013-07-22
影响因子：
4.7
作者：
Gero, Shane;Gordon, Jonathan;Whitehead, Hal
通讯作者：
Whitehead, Hal
影响因子：
2.5
作者：
Hobson, Elizabeth A.;Avery, Michael L.;Wright, Timothy F.
通讯作者：
Wright, Timothy F.