A0692
Title: Analyzing community ecology metabarcoding data using variance partitioning methods
Authors: Massimo Ventrucci - Department of Statistical Sciences, University of Bologna (Italy) [presenting]
Maria Franco Villoria - University of Modena and Reggio Emilia (Italy)
Luisa Ferrari - University of Modena & Reggio Emilia (Italy)
Alex Laini - University of Turin (Italy)
Abstract: Community ecology is an exciting field where ecologists and statisticians, in a collective effort, produce tools to investigate the distribution of species over space and the relationship these species have with environmental covariates. Generalized linear mixed models (GLMM) are pervasive in this field as they allow inferring both niche-related processes and dispersal. GLMMs are applied to a specific type of community ecology data called metabarcoding. Metabarcoding is a genetic technique that facilitates species identification through a sequence of operations: preprocessing samples collected in the field, targeting a small region of DNA, and sequencing. This produces large datasets where each sample has an attached label indicating a haplotype, i.e. a variant of a genetic species. One main goal is to separate the variance explained by the environmental covariates (abiotic factors) from the variance explained by the random effects (biotic factors such as dispersal and interaction). Several indicators to separate the contribution of fixed and random effects are proposed in the community ecology literature. These methods are discussed and compared with a novel procedure to perform variance partitioning based on, first, redefining the GLMM so that each model component is represented by a meaningful variance parameter and, second, posterior inference is derived reflecting the explained variance of each term. The method is illustrated with data from a real case study.
