Table 1.

Comparison of core subsets selected by MSTRAT, D-Method and Core Hunter

Strategy	MR	CE	SH	HE	NE	PN	CV
	Bulk data set
Core Hunter (single)†	0.572	0.641	4.531	0.667	3.446	0.000	100.000
Core Hunter (multi)‡	0.506	0.598	4.513	0.662	3.403	0.015	98.500
MSTRAT	0.477	0.571	4.493	0.649	3.217	0.021	97.900
D-Method§	0.503	0.578	4.411	0.626	2.980	0.066	93.400
COLLECTION	0.440	0.521	4.399	0.620	2.937	0.000	100.000
	Accession data set
Core Hunter (single)†	0.694	0.752	4.670	0.676	3.501	0.000	100.000
Core Hunter (multi)‡	0.659	0.733	4.613	0.650	3.281	0.084	91.600
MSTRAT	0.647	0.718	4.579	0.624	2.982	0.000	100.000
D-Method§	0.653	0.719	4.525	0.619	2.963	0.164	83.600
COLLECTION	0.630	0.696	4.467	0.591	2.742	0.000	100.000
	Population data set
Core Hunter (single)†	0.442	0.540	4.503	0.619	2.997	0.177	82.300
Core Hunter (multi)‡	0.396	0.508	4.482	0.609	2.969	0.225	77.500
MSTRAT	0.357	0.465	4.450	0.593	2.763	0.183	81.700
D-Method§	0.377	0.485	4.409	0.579	2.702	0.264	73.600
COLLECTION	0.357	0.455	4.466	0.592	2.749	0.000	100.000

†each selection criteria was attempted to be optimized independently by performing 20 runs with 100% weight given to the respective selection criteria during each run. Results reported for each measure are independent of results reported for all other measures.

‡20 independent runs were performed with equal weight given to each of the seven measures in an attempt to maximize (minimize) all measures simultaneously.

§for each measure, results are shown for the best performing strategy as reported in [9].