Modelling multivariate data using product copulas and minimum distance estimators: an exemplary application to ecological traits

Modelling and applying multivariate distributions is an important topic in ecology. In particular in plant ecology, the multidimensional nature of plant traits comes with challenges such as wide ranges in observations as well as correlations between several characteristics. In other disciplines (e.g., finances and economics), copulas have been proven as a valuable tool for modelling multivariate distributions. However, applications in ecology are still rarely used. Here, we present a copula-based methodology of fitting multivariate distributions to ecological data. We used product copula models to fit multidimensional plant traits, on example of observations from the global trait database TRY. The fitting procedure is split into two parts: fitting the marginal distributions and fitting the copula. We found that product copulas are well suited to model ecological data as they have the advantage of being asymmetric (similar to the observed data). Challenges in the fitting were mainly addressed to limited amount of data. In view of growing global databases, we conclude that copula modelling provides a great potential for ecological modelling.

Receiver operating characteristic (ROC) curves: equivalences, beta model, and minimum distance estimation

Receiver operating characteristic (ROC) curves are used ubiquitously to evaluate scores, features, covariates or markers as potential predictors in binary problems. We characterize ROC curves from a probabilistic perspective and establish an equivalence between ROC curves and cumulative distribution functions (CDFs). These results support a subtle shift of paradigms in the statistical modelling of ROC curves, which we view as curve fitting. We propose the flexible two-parameter beta family for fitting CDFs to empirical ROC curves and derive the large sample distribution of minimum distance estimators in general parametric settings. In a range of empirical examples the beta family fits better than the classical binormal model, particularly under the vital constraint of the fitted curve being concave.

Adversarial Inference Is Efficient

We study properties of the adversarial framework, introduced in Kaji, Manresa and Pouliot (2020). We show that the adversarial inference with an oracle classifier is statistically efficient. In addition, we study the finite sample properties of the adversarial estimation framework for the autoregressive parameter of a linear dynamic fixed effects panel data model with Gaussian errors. Unlike maximum likelihood, but similarly as other minimum distance estimators, the adversarial estimators do not suffer from the incidental parameter bias. In our simulations, using a one-hidden-layer neural network as discriminator delivers the estimates with smallest root mean squared error.