Table 4.

Application of transformation method in a single train-test split to determine reduction in batch-related difference in measurements (Step 3).

	Train (N)	Rsq	Beta	Itc	Test (N)	ICC	Mean diff before	BA before sig.	Mean diff after	BA after sig.	Delta abs- diff	Delta abs-diff sig.
1× train-test split FTLD-Tau
GM	12	0.92	0.91	1.05	27	0.96	–1.16	5.04e-13	–0.05	0.513	–0.86	2.89e-11
WM	12	0.97	0.86	1.03	27	0.95	–1.32	2.08e-11	–0.03	0.808	–0.79	4.92e-07
1× train-test split FTLD-TDP
GM	24	0.70	1.01	0.89	29	0.88	–1.02	1.48e-08	–0.16	0.226	–0.61	1.92e-08
WM	24	0.75	0.79	–0.08	29	0.88	–1.02	4.37e-08	–0.20	0.117	–0.50	1.31e-04

BA, Bland-Altman statistics; Delta abs-diff, change in absolute difference; diff, difference; FTLD-Tau, frontotemporal lobar degeneration with inclusions of the tau protein; FTLD-TDP, frontotemporal lobar degeneration with inclusions of the transactive response DNA-binding protein 43 kDa; ICC, intraclass correlation coefficient; Itc, intercept; GM, gray matter; N, number of tissue samples; Rsq, R squared; sig., significance; WM, white matter. Here, we display the results of the application of our cross-validated transformation method in a single randomly obtained train-test split. On the left side, table shows transformation prerequisites (i.e., Rsq) and equivalence factors (i.e., beta, intercept) in training sets (GM/WM in FTLD-Tau and FTLD-TDP). On the right side, we report corresponding transformation outcomes (i.e., ICC) in the complementary testing sets. Additionally, we look at measures of test-retest agreement, i.e., mean difference between SB1 and SB2 before and after the transformation, with Bland-Altman statistics (p < 0.05, significant mean difference between SB1 and SB2/t-SB2; p > 0.05, non-significant mean difference between SB1 and SB2/t-SB2), as well as the decrease in absolute difference tested with a one-sample t-test statistics (p < 0.05, significant reduction in difference between duplicate measurements after the transformation).