抽样方式对社会网络分析的影响研究

Evaluating Influence of Sampling Methods Upon Social Network Analysis

社会网络分析（SNA）是一种从定量角度研究分析对象网络的关系结构及其属性的方法，近年来在诸多领域得到了应用。现有的研究多注重分析SNA各项指标所带来的解释意义，然而在具体的实证研究中，对于抽样方式的选择还不够重视，这在一定程度上会影响研究结论的准确性和有效性。本研究选取了抽样技术中最为经典的四种方式（节点抽样、边抽样、深度优先抽样和滚雪球抽样）作为分析对象，在无标度网络数据集下评估其对于SNA五种典型指标，即直径、平均距离、聚类系数、点度中心性和中介中心性的影响情况。本文的结论对SNA研究过程的规范性具有重要意义。

Social Network Analysis（ SNA）, a quantitative research method which aims at analyzing the structure and various properties of target network, has been applied in many academic fields. However, those existing studies emphasize more on the explanations of SNA indexes, rather than evaluating the impact of process elements to the research conclusion. This study selected four classic sampling methods （node sampling, link sampling, depth-first sampling and snowball sampling） as the research object, and briefly analyzed the impact of different sampling methods to five typical SNA indexes （diameter, average distance, clustering coefficient, degree centrality and betweenness centrality） in scale-free networks, which can be helpful to formalize the research process of implementing SNA.