Figure 2. Predictive power obtained for Bacteria, Archaea, Fungi, and
Protist and evaluated using the adjusted maxTSS for ‘presence-absence’
models (a,b,c,d), and Spearman’s rho (⍴) for ‘relative-abundance’ models
(e,f,g,h). For each threshold, the proportion of individual OTU models
that obtained a greater value is shown. The x-axis labels are explained
as follows: Fitted: the algorithm was able to fit a model;
TSS>TSSnull: the predictive power of a
model evaluated against null models; TSSadj: predictive
power metric rescaled so that TSSadj=0 corresponds to a
model with a predictive power equal to the 5% best null models.
Bacteria and archaea models were fitted with 250 sites, while fungi were
fitted with 217 and protists with 166 sites.