Опубликована: Дек. 10, 2024
Objective: With the aim of improving monitoring reliability and interpretability CI DM experimental statistical tests, we evaluate performance cutting-edge nonparametric tests post hoc procedures. Methods: A Friedman Aligned Ranks test, Quade multiple corrections Bonferroni-Dunn Holm were used to comparative analyze data. These approaches employed algorithm metrics with varied datasets their capability detect meaningful differences control Type I errors.Results: Advanced methods consistently outperformed traditional parametric offering robust results in heterogeneous datasets. The test was most powerful stable, procedures greatly increased power pairwise comparisons.Novelty: We advanced experiments: (Bonferroni-Dunn Holm). represent a departure from that depend on assumptions normality homogeneity variance, allowing for more flexible analyses complex, By comparing strength efficacy these methods, research also delivers common guidelines use; as well demonstrating utility realistic situations characterized by non-standard dispersed data.Implications Research: findings have far-reaching theoretical pragmatic implications scholars DM. On level, this work undermines bias towards techniques, providing an increasingly framework analysis research. This improves understanding adaptation fit complexities real-world data highlighting advantages specifically corrections. Practical give owners summaries actionable recommendations, which will assist researchers selection are tuned nature datasets, resulting improved future evaluations algorithms. Thus, endeavor promote statistically appropriate studies, leading confident valid claims surrounding algorithmic performance.
Язык: Английский