Statistical data representation in the context of increasing construction efficiency

Abstract

The purpose of the paper is the development of new statistical tests and new forms of their representation which add classic tests and classic form of their description. The development of new building materials requires several years and is quite expensive. Builders cannot enter the market and recoup their investments because of the need to test a new matter for sufficiently large samples of buildings. The tests in terms of time and cost resources can be comparable to the ones for development of a new substance. Thus, the experiments can be very expensive, which makes it difficult to quickly get enough statistical data to increase construction efficiency. That is why it is important to develop new statistical tests and new forms of their representation. It is shown that the spectrum of a geometric mean molecule has a smaller number of spectral lines for small samples of biometric data as compared with the molecule of chi-square criterion. In addition, the probability amplitudes of the spectral components of a geometric mean molecule have a close to normal distribution. This advisably distinguishes the geometric mean molecule from the chi-square molecule. The criterion for the geometric mean of small samples has greater power as compared with the chi-square criterion. In order to increase the accuracy of statistical analysis of small samples, it is necessary to extend a variety of statistical criteria both in the usual continuous representation of their states, and in the analysis of their discrete spectra of the probability amplitudes. To increase the efficiency of the construction sector, it is necessary to transfer to the use of deep neural networks that analyze dozens of new and old statistical criteria, their equivalent neurons and equivalent artificial molecules. Technological capabilities for creating these structures already exist today.