“…Note that we may have chosen not to implement step (ii) above, or perhaps, even though step (ii) may reduce dimensionality dramatically ( m ≪ p ), we may still have N < m , or m may begin to approach N such that some form of regularization is still desirable. In such cases, as we are using Gaussian copulas, a variety of methods for regularizing the inverse covariance matrix may be considered (Friedman, Hastie, & Tibshirani, ; Schäfer & Strimmer, ; Ullah & Jones, ; Yuan & Lin, ); for simplicity, we shall not pursue the topic of regularization further here. Furthermore, we hasten to add that neither rare species nor unassociated species are omitted from the copula models that follow, but they are presumed to be independent of other species.…”