Fig. 3From: Difficulty in inferring microbial community structure based on co-occurrence network approachesRelationships between co-occurrence network performance (AUPR value) and network size n according to the network types: random networks (random), scale-free networks (sf), and small-world networks (sw). Random interaction matrices were considered. The cases of sparse networks (〈k〉 = 2; top panels) and dense networks (〈k〉 = 8; bottom panels) are shown. As representative examples, Pearson’s correlation-based method (a classical correlation-based method; a and b), Pearson’s partial correlation-based method (a classical graphical model-based method; c and d), CCLasso (a correlation-based method for compositional data; e and f), and SPEIC-EASI (a graphical model-based method for compositional data; g and h) are shown. smax was set to 0.5. The number of samples was set to 300Back to article page