On Nonparanormal Likelihoods

TL;DR

This paper introduces novel nonparanormal likelihood functions for flexible multivariate modeling, addressing the estimation challenges and proposing computational methods, with applications in discriminant analysis and correlation estimation.

Contribution

It develops four new nonparanormal log-likelihood functions, discusses their optimization properties, and provides computational techniques for improved estimation in semiparametric models.

Findings

Introduction of four nonparanormal likelihood functions

Discussion of convexity and optimization strategies

Application to discriminant analysis and correlation estimation

Abstract

Nonparanormal models describe the joint distribution of multivariate responses via latent Gaussian, and thus parametric, copulae while allowing flexible nonparametric marginals. Some aspects of such distributions, for example conditional independence, are formulated parametrically. Other features, such as marginal distributions, can be formulated non- or semiparametrically. Such models are attractive when multivariate normality is questionable. Most estimation procedures perform two steps, first estimating the nonparametric part. The copula parameters come second, treating the marginal estimates as known. This is sufficient for some applications. For other applications, e.g. when a semiparametric margin features parameters of interest or when standard errors are important, a simultaneous estimation of all parameters might be more advantageous. We present suitable parameterisations of nonparanormal models, possibly including semiparametric effects, and define four novel nonparanormal log-likelihood functions. In general, the corresponding one-step optimization problems are shown to be non-convex. In some cases, however, biconvex problems emerge. Several convex approximations are discussed. From a low-level computational point of view, the core contribution is the score function for multivariate normal log-probabilities computed via Genz' procedure. We present transformation discriminant analysis when some biomarkers are subject to limit-of-detection problems as an application and illustrate possible empirical gains in semiparametric efficient polychoric correlation analysis.