New perspectives in the measurement of biodiversity

Entropy is widely used in biodiversity studies, where data often present complex interactions. Difficulties arise in linking entropy to available covariates or data dependence structures, as all existing entropy estimators assume independence. We take a Bayesian model-based approach and focus on estimating the probabilities which compose an entropy index, accounting for data dependence. This way, the entropy estimate is not a single value, rather it becomes a curve or a two dimensional surface according to the data structure. We obtain an interpretable index of the latent biodiversity of a system.