倾向得分匹配法:促进数据科学视角下情报学研究的因果推断

Propensity Score Matching:Facilitating the Causal Inference of Data Science-Oriented Information Studies

在大数据时代,数据科学驱动的情报学领域的量化研究亟须从相关分析延伸到因果分析,从而更有效地挖掘数据价值并拓展情报学量化分析的新方法。本文为情报学领域引入一种基于二手数据或者观测数据进行处理效应分析的统计方法:倾向得分匹配法。首先,在厘清倾向得分匹配法的来源与原理的基础上,本文介绍了该方法的具体实施步骤。接着,探讨了数据驱动型情报学量化研究领域对对照实验和因果推断的需求,强调了倾向得分匹配法的优势和革新、价值与贡献。然后,基于对倾向得分匹配法的文献检索,归纳了其在情报学及相关领域的应用现状,阐述了该方法在情报学的适用研究方向。最后,本文展望了该方法在情报学领域的应用前景与未来挑战,以期为数据科学视角下情报学领域定量研究的方法借鉴和革新提供新思路。

In the era of big data,data science-oriented quantitative research in the information science field must go beyond analysis of correlation to analysis of causality,and more effectively mine data value and expand new quantitative methods in information studies.This article introduces a statistical method for measuring treatment effects based on secondhand or observational data,called propensity score matching(PSM).First,we introduce the implementation steps of PSM to clarify the method’s origins and principles.Second,we discuss the need for controlled experiments and causal inference in the field of data science-oriented quantitative information studies,and emphasize the advantages and contributions of PSM.Third,we conduct an in-depth review of existing work employing the PSM approach in both information science and related fields,and elaborate upon directions for future research applying PSM in information science.Finally,we describe the future prospects and challenges of this method in the field of information science.This study thus sheds light on the quantitative area of information science from a data science perspective.