As in Figure 1, but in this run, splitting of real test data DATA-I is undertaken to maximise the value S_ρ of the inner product S_ρ of item score vectors in the g-th and h-th subtests. Here, the lower right panel displays a plot against the splitting index t, of the value S² of the absolute difference between sum of squares of components of the item score vectors in the current g-th and the current h-th subtests. N.B. Due to the permitted swapping of the j-th item of the current g-th subtest by the j′-th item of the current h-th subtest, (j ≠ j′), under splitting by maximisation of S_ρ, sum of components of the 2 subtest item score vectors, can be more different, than when swapping across rows of the 2 subtests is not permitted, as under splitting by minimising S.