Multivariate quantiles and multiple-output regression quantiles: From $L_1$ optimization to halfspace depth

Abstract

A new multivariate concept of quantile, based on a directional version of Koenker and Bassett's traditional regression quantiles, is introduced for multivariate location and multiple-output regression problems. In their empirical version, those quantiles can be computed efficiently via linear programming techniques. Consistency, Bahadur representation and asymptotic normality results are established. Most importantly, the contours generated by those quantiles are shown to coincide with the classical halfspace depth contours associated with the name of Tukey. This relation does not only allow for efficient depth contour computations by means of parametric linear programming, but also for transferring from the quantile to the depth universe such asymptotic results as Bahadur representations. Finally, linear programming duality opens the way to promising developments in depth-related multivariate rank-based inference.

This paper discussed in: [arXiv:1002.4494], [arXiv:1002.4496], [arXiv:1002.4509]. Rejoinder in [arXiv:1002.4515]. Published in at http://dx.doi.org/10.1214/09-AOS723 the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (1)

• Quantile tomography: using quantiles with multivariate data

Cited by in corpus (30)

• A review of predictive uncertainty estimation with machine learning
• Multivariate and functional classification using depth and distance
• A Directional Multivariate Value at Risk
• On multivariate quantiles under partial orders
• Local bilinear multiple-output quantile/depth regression
• Gibbs posterior inference on multivariate quantiles
• Bridging centrality and extremity: Refining empirical data depth using extreme value statistics
• Finite sample breakdown point of Tukey's halfspace median
• A Bayesian Approach to Multiple-Output Quantile Regression
• Center-Outward Distribution Functions, Quantiles, Ranks, and Signs in ${\mathbb{R}}^d$
• Directional Multivariate Extremes in Environmental Phenomena
• Multivariate Conditional Transformation Models
• Fast computation of Tukey trimmed regions and median in dimension $p>2$
• Calibrated Multiple-Output Quantile Regression with Representation Learning
• Exact computation of the halfspace depth
• Tukey depth: linear programming and applications
• Fast implementation of the Tukey depth
• The Sparse Multivariate Method of Simulated Quantiles
• Vector quantile regression beyond correct specification
• Fast and exact implementation of 3-dimensional Tukey depth regions
• Discussion of "Multivariate quantiles and multiple-output regression quantiles: From $L_1$ optimization to halfspace depth"
• Multivariate Uncertainty Quantification with Tomographic Quantile Forests
• Quantile Mixed Hidden Markov Models for multivariate longitudinal data
• Set-valued conditional functionals of random sets
• Multivariate Geometric Expectiles
• Quantile Regression with Interval Data
• Comparing dominance of tennis' big three via multiple-output Bayesian quantile regression models
• Static and Dynamic Models for Multivariate Distribution Forecasts: Proper Scoring Rule Tests of Factor-Quantile vs. Multivariate GARCH Models
• Exactly computing bivariate projection depth contours and median
• A Pseudo-Metric between Probability Distributions based on Depth-Trimmed Regions