多源数据的期刊复合引文分析均值方法探析

Mean Algorithms for Composite Journal Citation Analysis Based on Multi-source Data

[目的/意义]鉴于Scopus、Web of Science(WoS)与Google Scholar(GS)3个单源数据库在计量学研究和应用中的局限,发展综合利用多源数据信息的复合型引文分析方法可为计量学分析提供一种补充思路。[方法/过程]以期刊总被引次数这一底层参量为切入点,用图书情报学国际期刊2009-2014年数据为实证基础,尝试对多源数据复合引文算术均值、几何均值及调和均值方法进行比较讨论。[结果/结论]结果发现,算术均值、几何均值和调和均值虽有差异,但亦有线性关联,应用中可选其一;累计算术均值、几何均值和调和均值与期刊数之间未证实存在布拉德福定律,但在期刊数量等同的三分区内,累计被引均值满足形如n^2:n:1的经验分布;多源数据复合引文与单个数据库引文的期刊排名差异呈现"两端小、中间大"之现象;期刊的算术均值排名与GS排名结果更接近,几何均值排名则与Scopus相似度更高,而调和均值与WoS的排名差异最小。

[ Purpose/significance] As three single-source databases, Scopus, Web of Science and Google Scholar all have limits in informetric researches and applications. The composite citation analysis method, which utilizes multi - source data, can provide a supplemental method. [ Method/process ] Based on the journal＇ s total citations and the cases of international LIS journals, this paper attempts to compare the mean algorithms of composite citation analysis, including arithmetic mean, geometric mean and harmonic mean. [ Result/conclusion] The results show that, though arithmetic mean is different from geometric mean or harmonic mean, there is a strong correlation between them. The cumulative arithmetic mean, cumulative geometric mean and harmonic mean do not follow Brad Ford＇ s law, but with the same number of journals in three sub-zone, the cumulative value satisfies the model of n2 ：n：1. The difference between journal rankings in single-source database and multi - source data appears as the phenomenon of balance in ends, difference in mid. The journal ranking of arithmetic mean is closer to GS, while geometric mean is closer to Scopus and harmonic mean is closer to WoS.