Table 12.

Integrated T-Test comparisons of model and statistical method performance (Monte Carlo simulations: 10,000 replications, n = 25–500; 10-fold CV on CICIDS2017/UNSW-NB15).

Comparison/Method	t-Statistic	df	p-Value	Cohen’s d	False Positive Reduction	Type I Error	Power	Std Error Reduction	References
Proposed Hybrid (bootstrap-kernel)	3.45 (overall) 4.12 (subgroup, n = 45)	38.2/45.0	< 0.001	0.76–1.12	18–22%	0.048 ± 0.002	0.92 (SD = 0.01)	18% vs. Yuen 22% vs. HC4	This study
vs. HC4	3.88	45.0	< 0.001	0.85	22%	0.056	0.80	—	Cribari-Neto (2004)33
vs. BCa Bootstrap	3.21	45.0	< 0.002	0.76	20%	< 0.05	0.81	—	Efron (1987)34, Chwialkowski et al. (2014)36
vs. Yuen’s Trimmed t	4.05	45.0	< 0.0001	0.90	18%	Higher variance	0.81	—	Yuen (1974)35, Osatohanmwen (2025)46