. 2016 Jan 26;11(1):e0147261. doi: 10.1371/journal.pone.0147261

Table 4. Performance across variations in sample size.

Sample size	Method	MAD	NOTFront	PU_20%	PO_20%	r_s
n = 20	DEA	0.241	36.4%	0.0%	80.0%	0.747
	rDEA	0.123	7.5%	27.5%	33.0%	0.782
	rSDF-CD	0.149	0.0%	27.9%	40.3%	0.656
	ENS	0.102	0.0%	20.7%	34.7%	0.793
n = 100	DEA	0.105	16.4%	0.0%	55.4%	0.837
	rDEA	0.039	3.3%	6.1%	13.8%	0.936
	rSDF-CD	0.111	0.0%	49.3%	13.3%	0.748
	ENS	0.059	0.0%	26.4%	9.1%	0.921
n = 200	DEA	0.068	11.8%	0.0%	42.6%	0.863
	rDEA	0.025	2.7%	1.5%	7.2%	0.955
	rSDF-CD	0.106	0.0%	50.2%	10.5%	0.762
	ENS	0.055	0.0%	25.8%	6.3%	0.936
n = 1,000	DEA	0.017	5.8%	0.0%	20.5%	0.907
	rDEA	0.007	2.6%	0.0%	2.8%	0.963
	rSDF-CD	0.099	0.0%	47.2%	10.2%	0.774
	ENS	0.051	0.0%	23.7%	5.5%	0.938

Note: Numbers in bold highlight the best outcome for each performance indicator across the alternative approaches. MAD: median absolute deviation, NOTFront: percentage of misclassified DMUs, PU_20%: percentage of underestimation, PO_20%: percentage of overestimation, r_s: Spearman’s rank correlation.