arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1507.03408physics.soc-ph
收藏

自然分布和社会分布的分类第一部分:描述

A classification of natural and social distributions Part one: the descriptions

L. Benguigui, M. Marinov

摘要:

本文对自然科学和社会科学中的正则分布作了广泛的研究。该调查包括来自广泛学科的研究,以创建用于描述分布的不同数学函数的清单。该研究的目的是确定是否可以使用一个唯一的函数来描述所有的分布(普遍性),或者一个特定的函数最适合描述每个特定研究领域的分布(域普遍性)。我们提出了一个分类的分布到第八个不同的类别,基于三个表示:Zipf表示,累积密度函数(CDF)和概率密度函数(PDF)。在调查的89个案例中,既没有发现普适性,也没有发现域普适性。特别是,根据调查结果,“幂律为大多数分布提供了良好的描述”的说法可能会被拒绝。在我们的调查中,只有三分之一的分布与幂律相关,而另外三分之一的分布可以用对数正态分布和类似的函数(Dagum,Weibull,loglogistic和Gamma函数)来描述。我们认为,分布的正确表征依赖于两个条件。首先,重要的是包括可用数据的全部范围,以避免由于任意截止值而导致的失真。其次,建议以所有三种表示形式显示数据:Zipf表示,CDF和PDF。

英文摘要:

This paper presents an extensive survey of regular distributions in natural and social sciences. The survey includes studies from a wide scope of academic disciplines, in order to create an inventory of the different mathematical functions used to describe the distributions. The goal of the study was to determine, whether a unique function can be used to describe all the distributions (universality) or a particular function is best suited to describe the distributions in each specific field of research (domain universality). We propose a classification of distributions into eighth different categories, based on three representations: the Zipf representation, the cumulative density function (CDF) and the probability density function (PDF). In the 89 cases included in the survey, neither universality nor domain universality was found. In particular, based on the results of the survey, the claim that "power law provides a good description for majority of distributions" may be rejected. Only one third of the distributions in our survey are associated with power laws, while another third is well described by lognormal and similar functions (Dagum, Weibull, loglogistic and Gamma functions). We suggest that correct characterization of a distribution relies on two conditions. First, it is important to include the full range of the available data to avoid distortion due to arbitrary cut off values. Second, it is advisable to display the data in all three representations: the Zipf representation, the CDF and the PDF.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑