Handcock Mark S, Morris Martina
Department of Statistics, University of Washington, Seattle, USA.
Comput Math Organ Theory. 2009 Dec 1;15(4):294-302. doi: 10.1007/s10588-008-9055-x.
Networks are being increasingly used to represent relational data. As the patterns of relations tends to be complex, many probabilistic models have been proposed to capture the structural properties of the process that generated the networks. Two features of network phenomena not captured by the simplest models is the variation in the number of relations individual entities have and the clustering of their relations. In this paper we present a statistical model within the curved exponential family class that can represent both arbitrary degree distributions and an average clustering coefficient. We present two tunable parameterizations of the model and give their interpretation. We also present a Markov Chain Monte Carlo (MCMC) algorithm that can be used to generate networks from this model.
网络正越来越多地用于表示关系数据。由于关系模式往往很复杂，人们提出了许多概率模型来捕捉生成网络的过程的结构特性。最简单的模型未捕捉到的网络现象的两个特征是个体实体拥有的关系数量的变化及其关系的聚类。在本文中，我们提出了一个属于弯曲指数族类的统计模型，它可以表示任意度分布和平均聚类系数。我们给出了该模型的两种可调参数化方法并对其进行了解释。我们还提出了一种马尔可夫链蒙特卡罗（MCMC）算法，可用于从该模型生成网络。
