Computational protocol: Spatial, Phylogenetic, Environmental and Biological Components of Variation in Extinction Risk: A Case Study Using Banksia

Similar protocols

Protocol publication

[…] Freckleton & Jetz [] presented a generalized least-squares model that partitions the variance in a response variable into a linear combination of phylogenetic, spatial, and independent (neither phylogenetic nor spatial) variance components: V(ϕ,λ)=γh+λ′Σ+ϕWWhere Σ is a variance-covariance matrix describing the phylogenetic distances among species, W is a variance-covariance matrix describing spatial distances among species, and h is a vector of tip heights from the phylogeny. The parameter λ′ = (1 − ϕ)λ, where λ is Pagel’s measure of phylogenetic signal [], and represents the phylogenetic component. The parameter ϕ represents the spatial component, and γ = (1 − ϕ)(1 − λ) represents the independent component. The three parameters sum to one, and can be interpreted as the relative contribution of phylogenetic, spatial and independent effects, provided the phylogenetic and spatial matrices are scaled to the same units.We generated a phylogenetic matrix from the Banksia maximum clade credibility tree using the function “vcv.phylo” in the ape library for R. To generate the spatial matrix we used the function “earth.dist” in the fossil library for R to obtain great-circle distances between the centroids of species distributions. We used the function “regress” in the regress library for R [] to fit models in which the phylogenetic and spatial variance components are jointly estimated by maximum likelihood, and we then used these to calculate ϕ, λ’, and γ.To fit models we used a two-stage approach. For each response variable (threat status and range size), we first fitted a full model that included all predictor variables together with phylogenetic and spatial matrices, and simplified this to a minimum adequate model in which all predictors were significant. This identified the sets of factors associated with extinction risk independently of one another, and independently of phylogenetic and spatial effects. We then took these sets of predictors forward for further model comparison. We used the Akaike Information Criterion (AIC) to compare the fit of the full models (predictors + phylogeny + space) to models including predictors only, phylogeny + space only, predictors + phylogeny, and predictors + space. The R code used for the analysis is provided in . […]

Pipeline specifications

Software tools PHYSIG, APE
Application Phylogenetics
Organisms Homo sapiens