Fig. 1

Nestedness for NESTLON under variation of both exogenous parameters θ _con and θ _nest on a perfectly nested unipartite benchmark graph (i.e. α=0.49). All vertices belong to a single nested component and, therefore, we choose a parameter set for which NESTLON detects all of them. In the left panel the thresholds are too rigid for θ _con =1 and θ _nest ≥0.5. Thus, θ _con <1 and θ _nest <0.5 are reasonable detection thresholds for NESTLON. In the right panel we force NESTLON to start with a randomly selected vertex (in contrast to the highest degree) for the same benchmark graph. Although, component sizes differ independently from θ _con <1 (about half of all vertices are detected on average due to random starting position), we infer the same calibration (θ _con <1 and θ _nest <0.5). We conclude that NESTLON works most reliable if we start with the highest degree vertex on a degree-ordered graph