Русский
!

Conference publications

Abstracts

XXII conference

Impact of social factors on experimantal data statistics: crystal structural databases

Slovokhotov Y.L.

Department of Chemistry, Moscow State University, 1 Vorobyavy Gory, Bld. 3, 119991, Moscow, Russia

1 pp. (accepted)

Numerical parameters and mathematical modeling are used in many social sciences. However, reliability of measurements on social system as well as reproducibility of their results remain not quite certain. Multiple different factors, influencing social systems, complicates separation of the measured data into signal and noise. General possibility to minimize the latter in social sciences is not obvious; it needs a special analysis. To solve this problem, big sets of unequivocally measured data, impacted by social factors, are of great importance.

We have analyzed distributions of several quantitative parameters of crystals, reflecting fragments’ geometry and symmetry. Such parameters are reliably determined for ca. 900000 structures for pure crystalline compounds. Most of these compounds were obtained artificially in different forms of collaboration under some definite economical and social requests. So the structural subsets retrieved from computer structural databases are strongly influences by “human”, i.e. economical and social factors, indirectly reflected in their statistical properties.

We discuss typical distribution of quantitative data obtained for ‘non-living’ physical systems as well as for ‘living’ social ones. Big subsets of structural data (ca. 103 – 104 points) for crystalline substances resemble ‘social’ datasets in their statistical features, viz. non-gaussian distribution, heavy tails, clustering of points, etc. This observation reflects such typical features of social systems as their micro- and mesoscopic scale (in number of agents), non-stationary random process, and autocorrelations in time series. The socially influences datasets are often not suitable for routine methods of statistical treatment regardless high level of reliability and precision of their data. These sets, considered as ‘bad’ ones in the empirical statistical analysis, often become a subject of voluntary corrections that bias an objective information therein.



© 2004 Designed by Lyceum of Informational Technologies №1533