|
This article is cited in 1 scientific paper (total in 1 paper)
Control in Social and Economic Systems
Experimental studies on the test questions based on the histogram ratio
N. Kalmykova, Yu. V. Sidel'nikovab a V.A. Trapeznikov Institute of Control Sciences of RAS, Moscow
b Moscow Aviation Institute (State Technical University)
Abstract:
The article proposes a histogram ratio as a tool of an integral assessment of the degree of confidence of experts in their estimates. Herewith, the experts evaluate test questions presented in the form of a problem. Experimental studies of test questions based on this ratio are proposed as a tool for studying the phenomenon of over-confidence among experts. A hypothesis about the correlation between the degree of confidence of an expert in his assessment and a posteriori assessment of his accuracy is proposed and substantiated. Explanations of the factors that determine the correctness of the conclusions obtained in these studies are provided. Requirements for the form and type of test questions have been introduced. Also, hypotheses that characterize the dynamics of changes in the values of the histogram ratio from the sample size are proposed and experimentally substantiated, and their consistency is confirmed. The dynamics of the integral assessment of experts ' confidence in their assessments were studied. The factors on which the conclusions of experimental studies depend are revealed. It is shown that the histogram ratio is not only an integral assessment of the degree of confidence of experts in their estimates, but also an analogue of any discrete indicator of the error of a particular expert when evaluating a particular question.
Keywords:
expert appraisals, test question, histogram ratio.
Received: August 31, 2020 Published: September 30, 2020
Citation:
N. Kalmykov, Yu. V. Sidel'nikov, “Experimental studies on the test questions based on the histogram ratio”, UBS, 87 (2020), 101–130
Linking options:
https://www.mathnet.ru/eng/ubs1060 https://www.mathnet.ru/eng/ubs/v87/p101
|
Statistics & downloads: |
Abstract page: | 130 | Full-text PDF : | 53 | References: | 28 |
|