2.7 | Testing for a phylogenetic signal
Inferences made from phylogeny-informed null modeling approaches are predicated on the existence of a positive phylogenetic signal in species niche-space (Webb, Ackerly, McPeek, & Donoghue, 2002); the notion that closely related species possess similar suites of traits and occupy similar niches (Tucker, Davies, Cadotte, & Pearse, 2018). We tested for a phylogenetic signal with respect to abundance in the presence of sandwort using the R package phylosignal (Keck, Rimet, Bouchez, & Franc, 2016). To approximate association with a sandwort-based diet, we estimated the abundance averaged niche space of each ASV. Briefly, abundances were relativized to the total sequence count for each ASV across all samples. Relative abundances among horses with access to sandwort were multiplied by 1 and relative abundances among horses without sandwort access were multiplied by −1. The sum of these values within each ASV were assigned as a ‘niche-score’ for each ASV which varied continuously between −1 (ASV only present in horses with access to sandwort) and 1 (ASV only present in horses without access to sandwort). Sandwort was selected since: 1) different dietary components are expected to vary in their polysaccharide composition, thereby selecting for different microbial metabolic functions (Julliand & Grimm, 2017) and 2) sandwort presence was observed in preceding analyses to be an important correlate of phylogeny-informed and phylogeny-independent β-diversity.