Clinics (Sao Paulo). 2011;66(12):2099-104. doi: 10.1590/s1807-59322011001200016.
Impact Factors (IF) are widely used surrogates to evaluate single articles, in spite of known shortcomings imposed by cite distribution skewness. We quantify this asymmetry and propose a simple computer-based procedure for evaluating individual articles.
(a) Analysis of symmetry. Journals clustered around nine Impact Factor points were selected from the medical ''Subject Categories'' in Journal Citation Reports 2010. Citable items published in 2008 were retrieved and ranked by granted citations over the Jan/2008 - Jun/2011 period. Frequency distribution of cites, normalized cumulative cites and absolute cites/decile were determined for each journal cluster. (b) Positive Predictive Value. Three arbitrarily established evaluation classes were generated: LOW (1.3≤IF<2.6); MID: (2.6≤IF<3.9); HIGH: (IF≥3.9). Positive Predictive Value for journal clusters within each class range was estimated. (c) Continuously Variable Rating. An alternative evaluation procedure is proposed to allow the rating of individually published articles in comparison to all articles published in the same journal within the same year of publication. The general guiding lines for the construction of a totally dedicated software program are delineated.
Skewness followed the Pareto Distribution for (1<K<2). Observed Positive Predictive Values ranged from 24 - 43% for over 98% of the selected journals in the ISI database. Continuously Variable Rating is shown to be a simple computer based procedure capable of accurately providing a valid rating for each article within the journal and time frame in which it was published.
尽管影响因子（IF）存在引文分布偏态带来的已知缺陷，但其仍被广泛用作评价单篇论文的替代指标。我们量化了这种不对称性，并提出了一种简单的基于计算机的方法来评价单篇论文。
（a）分析对称性。从 2010 年期刊引文报告的医学“学科类别”中选择了九个 IF 点附近的期刊聚类。检索并按 2008 年 1 月至 2011 年 6 月期间授予的引文进行排序。确定了每个期刊聚类的引文频率分布、归一化累积引文和绝对引文/十分位数。（b）阳性预测值。生成了三个任意确定的评估类别：低（1.3≤IF<2.6）；中（2.6≤IF<3.9）；高（IF≥3.9）。估计了每个类别的期刊聚类的阳性预测值。（c）连续变量评分。提出了一种替代的评估方法，允许对当年发表的同一份期刊中的每篇单独发表的文章进行评分，并与该期刊中发表的所有文章进行比较。概述了构建专门软件程序的总体指导原则。
偏态符合帕累托分布（1<K<2）。在 ISI 数据库中，超过 98%的选定期刊的观察阳性预测值在 24%到 43%之间。连续变量评分被证明是一种简单的基于计算机的方法，能够准确地为期刊和发表时间范围内的每篇文章提供有效评分。
