Figure 1.
The different types of biological surrogates (red) and their targets (green).
(A) Higher-taxa, where a taxon (or taxa) at a higher taxonomic level acts as surrogates for taxa at lower levels, (B) cross-taxa surrogates, where a taxon (or taxa) acts as a surrogate for another taxon (or taxa) at the same taxonomic level, and (C) subset-taxa surrogates, where a particular taxon (or taxa) acts as a surrogate for the entire target community. See Table S2 for referenced examples of each type of biological surrogate.
Table 1.
Statistical methods and biodiversity metrics used in marine biological surrogacy studies.
Table 2.
Cross-tabulations of the number of tests for each combination of factor levels.
Table 3.
Higher-taxa surrogate predictive power (R2) as a function of the number of taxonomic steps between the surrogate and the target.
Figure 2.
Surrogate effectiveness defined by P, the proportion of tests concluding that surrogate predictions were non-random.
Prior distribution (circles), posterior distribution given an uninformative prior (analogous to the likelihood; squares), and posterior distribution given an informative prior (diamonds). Error bars depict the standard deviation of the prior or posterior. Asterisks indicate the best model according to the deviance information criterion. Factors include the marine habitat (Habitat), spatial scale (Scale), the statistical method used to assess surrogate performance (Method) and the type of surrogate (Type).
Figure 3.
Surrogate effectiveness defined by R2, the predictability of targets using surrogates.
Prior distribution (circles), posterior distribution given an uninformative prior (analogous to the likelihood; squares) and posterior distribution given an informative prior (diamonds). Error bars depict the standard deviation of the prior or posterior. Asterisks indicate the best model according to the deviance information criterion. Factors include the marine habitat (Habitat), spatial scale (Scale), the statistical method used to assess surrogate performance (Method) and the type of surrogate (Type).
Figure 4.
Posterior distributions (given an informative prior) of surrogate effectiveness defined as P.
Posterior distributions of P (i.e. the proportion of tests concluding that surrogate predictions are non-random) are given according to the marine habitat (Habitat), spatial scale (Scale), the statistical method used to assess surrogate performance (Method) and the type of surrogate (Type). Asterisks indicate models outperforming the null model. A Gaussian distribution with the mean and standard deviation of the posterior distribution was used to approximate posterior distributions.
Figure 5.
Posterior distributions (given an informative prior) of surrogate effectiveness defined as R2.
Posterior distributions of R2 (i.e. the surrogate predictive power) are given according to the marine habitat (Habitat), spatial scale (Scale), the statistical method used to assess surrogate performance (Method) and the type of surrogate (Type). Asterisks indicate models outperforming the null model. A Gaussian distribution with the mean and standard deviation of the posterior distribution was used to approximate posterior distributions.
Table 4.
Deviance information criterion (DIC) for models of surrogate effectiveness defined as the proportion of tests (P) concluding that surrogate predictions are non-random and as the surrogate predictive power (R2).