<<<  Back

2021, № 90

UDC: 332.365:004.652
GSNTI: 83.77.01

Tools for econometric modeling of agricultural production factors

The digitalization of society has led to the fact that today information on all types of activities of enterprises and organizations is regularly collected and accumulated in real time, and many reports are generated. All this contributes to understanding the need for data-driven management. It is necessary to expand the traditional tools of the economist (grouping, correlation and regression analysis, time series analysis) with modern methods of "understanding data", identifying cause-and-effect relationships for the formation of recommendations, forecasts and interpretation of data and models. The emergence of the "understanding" effect is directly related to the growth of the level of analytical culture of the population. We believe that an acceptable level of analytical culture that allows us to talk about the adoption of the idea of data-driven management lies between a complete rejection of the use of data and "dataism" (the absolutization of data in decision-making processes that excludes human participation). An important factor that made it possible to realize this is the trend of interpretability of machine learning models( Machine Learning - ML), which practically reduced complex ML models to linear ones (2017-2019), which are the leitmotif of applied statistics. Thus, the trend of machine learning, which appeared in the 1960s on the wave of the statistical idea of the bootstrap method (the idea of multiplying a sample to obtain stable estimates), returned to interpreted statistical models almost 60 years later (as another confirmation of large Kondratiev cycles is possible, but already in data analysis). Therefore, today it is necessary to expand the traditional tools of the economist (grouping, correlation and regression analysis, time series analysis) with modern methods of "understanding data", identifying cause-and-effect relationships for the formation of recommendations, forecasts, interpretation of data and models. The processing strategy for data-based management (data driven) is always individual and depends on the data structure, its quality, properties, and management goals. In this article, based on the data on the activities of agricultural organizations in the Krasnodar Territory for 2018, we demonstrate one of the possible ways to "understand the data" based on the use of modern methods that use the ideology of sample multiplication (bootstrap method) in the statistical package JASP: network analysis, principal component analysis, cluster analysis, and regularized regression.
Keywords: Understanding, data driven, analytical culture, data array, factors, cognitive map, centrality measure, clustering, component analysis, cluster analysis, determination index, regression.
DOI: 10.21515/1999-1703-90-11-20

Authors:

1. Katsko Igor Alexandrovich, DSc in Economics, professor; Department of Statistics and Applied Mathematics, Federal State Budgetary Educational Institution of Higher Education “I.T. Trubilin Kuban State Agrarian University”.
2. Lyakhovetsky Alexey Mikhailovich, PhD in Economics, professor; Department of Statistics and Applied Mathematics, Federal State Budgetary Educational Institution of Higher Education “I.T. Trubilin Kuban State Agrarian University”.
3. Pertsukhov Viktor Ivanovich, PhD in Economics, associate professor; Department of Statistics and Applied Mathematics, Federal State Budgetary Educational Institution of Higher Education “I.T. Trubilin Kuban State Agrarian University”.