A Note on Learning Dependence under Severe Uncertainty

Abstract

We propose two models, one continuous and one categorical, to learn about dependence between two random variables, given only limited joint observations, but assuming that the marginals are precisely known. The continuous model focuses on the Gaussian case, while the categorical model is generic. We illustrate the resulting statistical inferences on a simple example concerning the body mass index. Both methods can be extended easily to three or more random variables. © Springer International Publishing Switzerland 2014.