期刊h指数模型修正与实例验证

Adjusted h-Index for Journals and Its Empirical Test

简要分析了h指数模型的的发展,重点分析了Glanzel-Schubert期刊h指数模型优点与可能存在的问题,并通过实例统计分析证实了常数c的理论值为1,且篇均被量的幂被高估了,在此基础上提出了新的修正模型（h=N^（0.34）S^（0.58））。为了实例验证修正前后模型的准确性,使用统计学中解释力度（r^2）概念,验证结果表明,即使在Glanzel-Schubert模型中常数c调整至接近最佳值时,总体上其解释力度也不如修正后的模型佳,这说明修正后的模型既保持模型的简洁性,同时更加与期刊h指数变化规律相一致。另通过大量统计样本分析,得出统计样本的载文时间段与被引用时间段的时间差大小对估算模型有一定的影响,并有一定的规律性。

This paper illustrates h-index model development briefly,and analyzes the advantage of Glanzel-Schubert h-index model as well as its existing problems.An empirical test verifies that the theoretical value of c is one and the times of per article cited are overestimated in this model.Based on this cognition,this paper proposes an adjusted model（h = N^0.34 S^0.58 ）.To test the correctness of the model,it quotes the concept of r^2 in statistics,and the result indicates that even though adjusted c is approximate to optimal value in Glanzel-Schubert model,its r^2 is not more than that in the adjusted model, which shows the adjusted model offers better simplicity and consistency with the change regularity of h-index.A statistical analysis of extensive sampling manifests that the time interval between the time of articles published and cited has certain influence on the estimation of model and appears certain regularities.