Comparison of the effect of mean-based method and z-score for field normalization of citations at the level of Web of Science subject categories

• Published in: Scientometrics Vol. 101; no. 3; pp. 1679 - 1693
• Main Authors: Zhang, Zhihui, Cheng, Ying, Liu, Nian Cai
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.12.2014; Springer
• Subjects: Bibliometrics. Scientometrics; Bibliometrics. Scientometrics. Evaluation; Citation analysis; Computer Science; Distributions; Exact sciences and technology; Impact analysis; Information and communication sciences; Information science. Documentation; Information Storage and Retrieval; Library and information science. General aspects; Library Science; Methods; Sciences and techniques of general use; Subject fields; Normalization effect; score; Mean-based method; Citation distribution; Field normalization; Research field; Correlation analysis; Impact factor; Scientific research; z-score; Citation; Comparative study
• ISSN: 0138-9130; 1588-2861
• DOI: 10.1007/s11192-014-1294-7

Field normalization is a necessary step in a fair cross-field comparison of citation impact. In practice, mean-based method ( m -score) is the most popular method for field normalization. However, considering that mean-based method only utilizes the central tendency of citation distribution in the normalization procedure and dispersion is also a significant characteristic, an open and important issue is whether alternative normalization methods which take both central tendency and variability into account perform better than mean-based method. With the aim of collapsing citation distributions of different fields into a universal distribution, this study compares the normalization effect of m- score and z -score based on 236 Web of Science (WoS) subject categories. The results show that both m -score and z -score have remarkable normalization effect as compared with raw citations, but neither of them can realize the ideal goal of “universality of citation distributions”. The results also suggest that m -score is generally preferable to z -score. The essential cause that m -score has an edge over z -score as a whole has a direct relationship with the characteristics of skewed citation distributions in which case m -score is more applicable than z -score.