Table 6.

U test statistic and p values as calculated for the differences between the model performance and explanation separability $S_{(a,b)}$ of the baseline and explanation ensemble models; a two-sided test was used.

Dataset (Task)	Model performance		Explanation consistency
	U Statistic	p value	U statistic	p value
BCW	75	0.00249292	774	0.009378
KAIMRC (Regression)	81	0.00040946	6475	$1.634\times {10}^{-6}$
KAIMRC (Classification)	51	0.04988344	3066	$6.382\times {10}^{-13}$
Codon usage (DNA)	81	0.00039825	5606.5	$3.855\times {10}^{-13}$
Codon usage (Kingdom)	0	0.00018267	11205	$1.179\times {10}^{-12}$
MIMIC-IV	72	0.10397974	1350	$8.226\times {10}^{-16}$