Partial correlations quantify linear association between two variables while adjusting for the influence of the remaining variables. They form the backbone for graphical models and are readily obtained from the inverse of the covariance matrix. For compositional data, the covariance structure is specified from log ratios of variables, which implies changes in the definition and interpretation o…