Table 2.

Rank correlations for each method with a gold standard obtained from the other half of the dataset

	1st split		2nd split		3rd split		4th split
method	1st half	2nd half	1st half	2nd half	1st half	2nd half	1st half	2nd half
GSZ-score	0.6201	0.6219	0.6421	0.6451	0.6594	0.6368	0.6176	0.653
t-test	0.5373	0.5276	0.5691	0.5748	0.5722	0.5827	0.6190	0.5727
KS test	0.4470	0.5089	0.4981	0.5140	0.5340	0.5054	0.4928	0.5388
modKS	0.5048	0.5772	0.5339	0.6035	0.5957	0.5336	0.5756	0.5873
iGA	0.5809	0.5888	0.6110	0.6133	0.6191	0.6148	0.5976	0.6185

Table shows rank correlation between the results for each half of a dataset with the gold standard ranking, obtained with all the methods from the other half (case ii evaluation). Various results are highlighted similarly to previous table. Two results for each split are obtained by testing the first half with a gold standard from the second half and testing the second half with a gold standard from the first half. Notice that GSZ-score clearly shows the best correlation. iGA is usually the second best method. The only exception is the first half of the 4th split, where t-test shows the best performance and GSZ-score scores as the close second best method.