基于自回归模型的社交网络依赖性检验

Testing of social network dependence based on the autoregressive model

随着信息技术的迅速发展,社交网络对人们生活的影响日益显著,对各种社交网络的分析和研究也显得越来越重要.识别和确定社交依赖性是社交网络分析的主要目标.本文提出一个改进的空间自回归模型,引入一个易感性指标对个体社交网络依赖性进行分类.基于本文所提出的模型,本文构造一个用于检验社交网络依赖是否存在的得分检验统计量,给出在原假设和备择假设下的检验统计量的渐近分布.当社交网络依赖性存在时,可以得到易感性指标的参数估计.本文通过模拟来评估所提出的检验统计量的经验表现.最后应用本文所提出的方法分析一个面向视频游戏的实时流媒体视频平台的数据集.

With the rapid development of information technology,the influence of social networks on people's lives is becoming increasingly significant,and the analysis and study of various social networks are becoming particularly important.Identifying and determining social dependencies are the main goal of social network analysis.In this paper,we propose an improved spatial autoregressive model that introduces a susceptibility index to classify individual social network dependence.Based on the proposed model,we construct a score test statistic to test the existence of social network dependencies and provide the asymptotic distribution of the test statistic under the null and alternative hypotheses.When social network dependence exists,the parameter estimation of the susceptibility index can be obtained.We conduct simulations to evaluate the empirical performance of the proposed test statistic.Finally,we apply the proposed method to analyze a dataset from a real-time streaming video platform for video games.