Komparativna analiza klasične inferencije i Bajesovog pristupa u obradi ekonomskih podataka

Abstract

Over the past century or more, one of the fundamental debates among statisticians has been about the meaning of the concept of probability. Essentially, there is a general consensus on the axioms of probability but its role in the process of reasoning is still questionable. It is where the separation into two basic approaches begins - classical and Bayesian. Bayesian analysis was dominant in the literature during the 19th century; however, in the 20th century classical statistics gained popularity due to the rigorous interpretation of probability. The last few years, interest in Bayesian methods has been growing again. The controversies around the application of these two approaches in the specific circumstances are still present today. It is the reason for the existence of two opposing schools. While some believe that classical methods show better performance in solving problems, others find the arguments in favor of Bayesian statistics. This duality contributes to the opinion that it is necessary to explore the possibilities that both approaches provide in the data processing. The subject of the doctoral dissertation research is the analysis of different methodological approaches of statistical inference and the possibility of their use in the processing of specific economic data. In order to respond to the subject of the study, this dissertation presents the historical development of two different approaches to inference, perceived characteristics of classical and Bayesian statistics, application possibilities of Bayesian methods in the processing of economic data, and finally, conclusions regarding the feasibility of the selected methods and models in economics. By evaluating the results of the applied maximum likelihood method- as the classical method- and Bayesian method on the example of one company, as well as a comparative analysis of the results obtained using both approaches in the analysis of trends of key financial indicators in the Republic of Serbia and some macroeconomic indicators in the European Union, the basic hypothesis that Bayesian approach to statistical inference has the advantage over classical inference was confirmed, as well as was the fact that its application in economics is justified.