, MODIFICATION OF THE FECHNER METHOD TO INCREASE THE ROBUSTNESS OF DATA ANALYSIS // BSU Bulletin. Mathematics, Informatics. - 2022. №1. . - С. 35-44.

MODIFICATION OF THE FECHNER METHOD TO INCREASE THE ROBUSTNESS OF DATA ANALYSIS

DOI: 10.18101/2304-5728-2022-1-35-44 UDK: 519.24

The article proposes a modification of the criterion for the consistency of the deviations of the values of the Fechner’s features by replacing the arithmetic mean with the median. Situations are given in which the use of the Fechner’s median crite- rion is expedient. The test example shows the difference mechanism for using the me- dian instead of the mean value when evaluating the correlation with this criterion. Based on the data on long-term observations of the results of athletes of different ages in the Irkutsk region, obtained by the authors in previous works, experiments were car- ried out with arrays of different sizes. The difference in the stability of the median Fechner method and the classical one is shown, both under conditions close to the nor- mal distribution and in the presence of an insignificant amount of outliers in the sample data. Including analyzed the behavior of the traditional and modified criteria for small samples.

consistency of sample data, Fechner’s law, Pearson correlation coeffi- cient, Student's t-test, average indicators.