Table 10.

Mean/standard deviation of the optimum loss scores for each true subcommunity under two different methods, with 10 simulations for each sample size

Sample size	Cluster	M1−M10	M9−M18	M17−M26
n=50	NMF	3/1.8	7.2/2	2.4/1.9
	Co-occurrence network	5.4/2.4	7.5/0.8	3.5/1.6
n=100	NMF	2.7/1.3	6.8/2.2	2.1/1.4
	Co-occurrence network	6.7/2.2	6/1.6	4.5/1.7
n=200	NMF	2.7/0.9	6.1/2	1.5/1.2
	Co-occurrence network	6.9/1.7	6.6/2.1	6.3/1.3