Comparative studies of preferential and weighted medians accuracy in robust estimation of central tendency; International Journal of Data Science and Analytics; Vol. 22

Comparative studies of preferential and weighted medians accuracy in robust estimation of central tendency; International Journal of Data Science and Analytics; Vol. 22

Bibliographic Details
Parent link:International Journal of Data Science and Analytics.— .— Cham: Springer Nature Switzerland
Vol. 22.— 2026.— Article number 262, 19 p.
Corporate Authors: Национальный исследовательский Томский политехнический университет Инженерная школа информационных технологий и робототехники, Национальный исследовательский Томский политехнический университет Инженерная школа Интеллектуальные энергетические систем
Other Authors: Muravyov (Murav’ev) S. V. Sergey Vasilyevich, Khudonogova L. I. Ludmila Igorevna, Shabramov V. S Vladislav Sergeevich, Ignatyev V. D. Vadim
Summary:A central tendency estimation for a given sample of some physical quantity values is a standard data analysis procedure. The data can be considered as intervals whose midpoints are the values and lower and upper boundaries are their uncertainties that are not necessarily equal to each other. The classical weighted median (WM) and the newly proposed preferential median (PM), estimators of central tendency, were experimentally tested and studied on a set of specially generated random normally and uniformly distributed synthetic input interval data. The WM is a 50% weighted percentile. The PM is the highest rank value in a consensus ranking found, using a Borda preference aggregation rule, for the set of rankings of discrete values corresponding to the input intervals. Both estimators are known to be robust. Absolute deviations from a given nominal value, standard deviations using bootstrapping and the deviations critical values at given standard confidence levels of both estimators were calculated. For normal input data, it was shown that accuracy, robustness and confidence of WM and PM are practically similar. For uniform input data, the PM is on average from 15 to 18% more accurate and from 17 to 21% more robust; the PM has lower critical values compared to the WM at all confidence levels, e.g., by absolute deviation, the PM superiority is from 0 to 55% at confidence level 0.95.
Текстовый файл
AM_Agreement
Language:English
Published: 2026
Subjects:
Central tendency estimator
Preferential median
Weighted median
Robustness
Interval fusion with preference aggregation (IF&PA)
электронный ресурс
труды учёных ТПУ
Online Access:https://doi.org/10.1007/s41060-026-01222-6
Format: Electronic Book Chapter
KOHA link:https://koha.lib.tpu.ru/cgi-bin/koha/opac-detail.pl?biblionumber=687843

Similar Items

Similar Items