Assessment of species abundance from metabarcoding data
To assess the level of variation in the number of reads explained by the proportion of a certain taxa in the sample, we calculated the proportion of reads or relative read abundance (RRA), the proportion of biomass and the proportion of individuals of every family in every plot. We used thebetareg function in the R package “betareg v1.1” (Cribari-Neto & Zeileis, 2010) to apply beta regression models to each family present in at least 10 plots. Two models were applied to each family, one for the RRA as a function of the percentage of mass the family represents in the sample, and one for the RRA as a function of the percentage of abundance. We applied the models to each family separately due to the non-independence of percentages in a sample, and we used 10 as the minimum presence in plots following the one-in-ten rule tested in other models (Peduzzi, Concato, Feinstein, & Holford, 1995). Beta regression models are designed for response variables with proportional data between 0 and 1. We calculated the adjusted R2 values and determined the global goodness of fit for each model. Only significant (p-value < 0.05) models with a pseudo-R2 > 0.5 and randomly distributed residuals were considered (Yellareddygari, Pasche, Taylor, Hua, & Gudmestad, 2015).
RESULTS