Ciminelli Joseph T, Love Tanzy, Wu Tong Tong
University of Rochester, 500 Joseph Wilson Boulevard, RC Box 270138, Rochester, NY 14623.
University of Rochester, 265 Crittenden Boulevard, Box 630, Rochester, NY 14642.
Spat Stat. 2019 Mar;29:129-144. doi: 10.1016/j.spasta.2018.11.001. Epub 2018 Nov 20.
Our work is motivated by a desire to incorporate the vast wealth of social network data into the framework of spatial models. We introduce a method for modeling the spatial correlations that exist over a social network. In particular, we model attributes measured for each member of the network as a continuous process over the social space created by their connections. Our method simultaneously models the unobserved locations of network members in social space and the spatial process that exists over that space based on the observed network connections and nodal attributes. The model is evaluated through simulation studies and applied to the importance ranking for a network of emergency response organizations and the physical activity habits of teenage girls. The introduced methods incorporate network data into the spatial framework, expanding traditional models to include this often relevant source of additional information.
我们的工作动机是希望将大量丰富的社交网络数据纳入空间模型框架。我们引入了一种对社交网络中存在的空间相关性进行建模的方法。具体而言，我们将为网络中每个成员测量的属性建模为基于其连接所创建的社交空间上的连续过程。我们的方法同时对社交空间中网络成员的未观测位置以及基于观测到的网络连接和节点属性在该空间上存在的空间过程进行建模。该模型通过模拟研究进行评估，并应用于应急响应组织网络的重要性排名以及少女的体育活动习惯。所引入的方法将网络数据纳入空间框架，扩展了传统模型，以纳入这一通常相关的额外信息来源。
