Taking seriously the arguments of Earman, Roberts and Smith that ceteris paribus laws have no semantics and cannot be tested, I suggest that ceteris paribus claims have a kind of formal pragmatics, and that at least some of them can be verified or refuted

For two-class discrimination, Ref. [1] claimed that, when covariance matrices of the two classes were unequal, a (class) unbalanced dataset had a negative effect on the performance of linear discriminant analysis (LDA). Through re-balancing 10 realworld datasets, Ref. [1] provided empirical evidence to support the claim using AUC (Area Under the receiver operating characteristic Curve) as the p...