banocc

"BAnOCC" (Bayesian Analysis of Compositional Covariance) is a Bayesian framework specifically designed to tackle the complexities of analyzing compositional data—data represented as vector proportions that sum to a constant total originating from unobserved counts. This sum constraint complicates the task of inferring correlations between features, a challenge particularly pronounced in ecology, where understanding such correlations is crucial.

BAnOCC innovatively employs a LASSO before estimating a sparse precision matrix, with the posterior distribution generated via Markov Chain Monte Carlo (MCMC) sampling. This approach not only facilitates the inference of the underlying network of correlations but also enables the quantification of uncertainty for any derived statistic of the precision matrix, including correlations. The framework employs a first-order Taylor expansion to approximate the transformation from unobserved counts to compositional data, offering insights into the factors that may influence the difficulty of inferring correlations.