COMPARATIVE ANALYSIS OF THE NORMALITY OF STATISTICAL CRITERIA FOR SAMPLES OF CONTAMINATED DATA
Journal of Rocket-Space Technology · 2025
DOI: 10.15421/452551 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Rocket-Space Technology, 2025. Verfasst von Oleksii Klymenko und 1 weiteren Autor:innen. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The paper considers the influence of contaminated samples with anomalous observations on the reliability of statistical analysis results and hypothesis testing for sample homogeneity. The main focus is on visual data analysis as an effective means of preliminary research. The use of histograms, scatter plots, and density estimates allows for the visual identification of outliers, the assessment of the distribution shape, and the detection of differences between samples. The purpose of the study is to evaluate the robustness of popular statistical criteria for testing the normality of distribution in the presence of contamination in small samples. The scientific novelty lies in the quantitative study of the impact of different types of contamination on the results of popular criteria, as well as in the practical assessment of their behavior under conditions of violation of assumptions about data homogeneity. The practical novelty lies in the development of recommendations for practitioners on the selection of the optimal criterion when analyzing samples with possible anomalous observations, taking into account the stability of statistical methods. Research methods include numerical modeling of samples with controlled introduction of structural contaminants, assessment of the frequency of false rejections/acceptance of the null hypothesis, as well as comparative analysis of the results obtained using the following statistical criteria: Student's t-test for comparing the mean values of two samples; the Kolmogorov-Smirnov one-sample test to verify the conformity of the empirical distribution with the theoretical one; the Anderson-Darling criterion to verify the normality assessment; the Kolmogorov-Smirnov two-sample test to verify the homogeneity of two distributions. The results of the study showed the importance of choosing the appropriate criterion depending on the sample size and the expected level of contamination. Presenting the average values and ranges for N repetitions of the experiment allows for a visual assessment of the stability and reliability of each test in the presence of contaminated data. Based on the experiments conducted, practical recommendations are proposed for the preliminary diagnosis of samples and the selection of the optimal approach to testing hypotheses in the presence of contaminated data.
Zitieren
APA
Klymenko, O., & Vovk, S. (2025). COMPARATIVE ANALYSIS OF THE NORMALITY OF STATISTICAL CRITERIA FOR SAMPLES OF CONTAMINATED DATA. Journal of Rocket-Space Technology. https://doi.org/10.15421/452551
Vancouver
Klymenko O, Vovk S. COMPARATIVE ANALYSIS OF THE NORMALITY OF STATISTICAL CRITERIA FOR SAMPLES OF CONTAMINATED DATA. Journal of Rocket-Space Technology. 2025. doi:10.15421/452551
BibTeX
@article{Klymenko2025,
author = {Klymenko, Oleksii and Vovk, Serhii},
title = {COMPARATIVE ANALYSIS OF THE NORMALITY OF STATISTICAL CRITERIA FOR SAMPLES OF CONTAMINATED DATA},
journal = {Journal of Rocket-Space Technology},
year = {2025},
doi = {10.15421/452551}
}
EUVIMED