随着教育与心理测量领域的发展，研究文献的日益增加，其对应的引文数据也形成了一个超大规模的复杂网络，因此识别该领域的发展现状和研究趋势就显得尤为重要。本文选取了教育与心理测量领域的6本权威期刊，分别为Psychometrika，British Journal of Mathematical and Statistical Psychology (BJMSP)，Journal of Educational and Behavioral Statistics (JEBS) ，Behavior Research Methods (BRM)，Psychological Methods，Structural Equation Modeling: A Multidisciplinary Journal，对发表在这六本期刊上且发表时间为2000年1月1日至2020年1月1日的文献利用python的爬虫技术爬取了标题，杂志名，发表时间，被引次数，关键词，作者，摘要，参考文献共8个基本标签，通过对脏数据的清洗，最终获得了4113份有效数据。

以4113篇文献构成的引文网络为目标，本文研究了引文网络的基本特点。首先，对作者的年生产能力以及每篇论文的年平均作者进行了统计，发现作者之间的合作关系不断加强。其次，为了确定最核心的论文，对每篇论文的引用量以及中心性指标进行了排序，得到以下结论：1、中心性分析结果靠前的论文与“缺失值”、“贝叶斯方法”以及“增长混合模型”这几个领域有关；2、“结构方程模型”领域在本次引文网络研究中更容易成为“传播中介”。最后，本文还对引文网络的入度分布进行了统计，发现了入度分布的偏态很严重，这间接表明了高入度的论文在引文网络中起着重要的作用。

为了进一步探索引文网络的社团结构，对比分析了FN算法和D-SCORE算法在引文网络中的社团检测结果。首先，两种算法均检测到了“聚类”、“在线实验”、“中介分析”、“词汇识别”以及“结构方程模型”社团，此外，D-SCORE算法还检测到了FN算法所未检测到的“情感”、“自适应测试”、“因果推断”等均非常重要的社团，从分类的合理性以及全面性来看，D-SCORE算法更胜一筹。为了探索教育与心理测量领域的研究主题随时间发展的趋势，本文还利用D-SCORE算法对社团的动态性质进行了研究，得到了以下结论：1、“结构方程模型”、“项目反应理论”以及“聚类”这三个领域最早具有社团属性；2、“聚类”社团以及“结构方程模型”社团发展较快；3、“在线实验”社团是近5年才被发现的社团，属于最新社团。

With the development of the field of psychological and educational measurement and the increasing number of research literature, the corresponding citation data has also formed a large scale complex network. It is particularly important to identify the development status and research trend of this field. In this paper, we choose six author -itative journals in psychological and educational measurement field, Psychometrika, British Journal of Mathematical and Statistical Psychology (BJMSP), Psychometrika, British Journal of Mathematical and Statistical Psychology Journal of Educational and Behavioral Statistics (JEBS), Behavior Research Methods (BRM), Psychological Methods, Structural Equation Modeling. We extract the title, journal name, public -cation date, citation times, keywords, author, abstract, and references for every article published in these six journals from January 1, 2000 to January 1, 2020 using Python's crawler technology. There are 8 basic labels for each article, and totally 4113 totally valid observations are obtained after clarifying dirty data.

Aiming at the citation network composed of 4113 references, this paper studies the basic characteristics of citation network.  First of all, the author's annual productivity and the annual average author of each paper are analyzed, and the cooperative relationship among authors is strengthened.  Secondly, in order to determine the most core papers, we rank the citations and centrality index of each paper, and draw the following conclusions: 1. The papers with higher centrality analysis results are related to the fields of "missing value", "Bayesian method" and "growth mixed model"; 2. The field of "structural equation model" is more likely to become a "communication intermediary" in this citation network study.  Finally, the paper also makes statistics on the input distribution of citation network, and finds that the skewness of the input distribution is very serious, which indirectly indicates that papers with high input play an important role in citation network.  

In order to further explore the community structure of citation network, the community detection results of FN algorithm and D-SCORE algorithm in citation network are compared and analyzed.  First of all, both algorithms detected "clustering", "online experiment", "mediation analysis", "word recognition" and "structural equation model" communities. In addition, D-SCORE algorithm also detected "emotion", "adaptive testing", "causal inference" and other very important communities, which were not detected by FN algorithm.  D-SCORE algorithm is superior in classification rationality and comprehensiveness.  In order to explore the development trend of educational and psychometric research topics over time, this paper also uses D-SCORE algorithm to study the dynamic nature of community, and draws the following conclusions: 1. "structural equation model", "item response theory" and "clustering" are the first three fields with community attributes; 2. "clustering" community and "structural equation model" community develop rapidly;  3. the "online experiment" club is the latest club that was discovered in the last five years.