期刊论文被引频次分布拟合与分析

Curve Fitting and Analysis on the Distribution of Citations to Journal Papers

以中国知网《中国学术期刊网络出版总库（CAJD）》为统计源，对不同学科与时间段样本的统计结果表明，论文被引频次分布曲线与布拉德福文献分布曲线基本一致。选用莱姆库勒公式进行曲线拟合，但统计样本中均有约0．5％的高被引论文的被引频次高于正常的曲线分布，同时在半对数曲线中也普遍存在“格罗斯下垂”，下垂误差为6％～7％。为了能够精确地得到拟合曲线，拟合时截除0．5％的高被引论文，同时对莱姆库勒公式加入修正系数，以修正“格罗斯下垂”误差。通过处理后，所有统计样本拟合均通过K—s检验。在具体分析各样本的基础上得出布拉德福系数k值具有学科特性，在论文发表7～8年内各学科k值存在明显的增长过程，k值的大小与删除0．5％的高被引论文后的最大被引频次（mm）、被引论文的平均被引频次（c）高度相关，并拟合出了三者之间的函数关系。

Using data from ＂ China Academic Periodicals Publication Library （CAJD）＂, statistical results from samples belonging to different subjects and time periods manifest the curve of cited distribution of papers and Bradford distribution is basic consistent , curve fitting use Leimkuble＇ s Formula, but the statistical samples are all have about 0. 5 % of highly cited papers, number of cited above the normal curve of distribution, while Groos droop is generally exist in semi- logarithmic curve, ＂droop＂ errors of about 6% -7%. In order to get accurate fitting curve , fitting to delete 0.5% of the highly cited papers, and adding Leimkuble＇ s Formula correction factor to amend ＂Groos droop＂ errors, all the fitting of statistical samples pass the KS test after processing. On the basis of the specific analysis of the sample＇ s, it obtained that Bradford coefficient （k） has the subject characteristics, k values of the various disciplines have obvious growth process in about 7 - 8 years after the papers published, k value has a high degree of correlation with the highest number of cited of papers after deleting the 0.5% highly cited papers （ m＇m ） and the average cited paper citations （c） , and fit out of a function between the three.