Predictive density estimation under the Wasserstein loss

Takeru Matsuda; William E. Strawderman

doi:10.1016/j.jspi.2020.05.005

Predictive density estimation under the Wasserstein loss

Research output: Contribution to journal › Article

Abstract

We investigate predictive density estimation under the L² Wasserstein loss for location families and location-scale families. We show that plug-in densities form a complete class and that the Bayesian predictive density is given by the plug-in density with the posterior mean of the location and scale parameters. We provide Bayesian predictive densities that dominate the best equivariant one in normal models. Simulation results are also presented.

Original language	English (US)
Pages (from-to)	53-63
Number of pages	11
Journal	Journal of Statistical Planning and Inference
Volume	210
DOIs	https://doi.org/10.1016/j.jspi.2020.05.005
State	Published - Jan 2021
Externally published	Yes

All Science Journal Classification (ASJC) codes

Statistics and Probability
Statistics, Probability and Uncertainty
Applied Mathematics

Keywords

Optimal transport
Predictive density
Shrinkage estimation
Wasserstein distance

Access to Document

10.1016/j.jspi.2020.05.005

Fingerprint Dive into the research topics of 'Predictive density estimation under the Wasserstein loss'. Together they form a unique fingerprint.

Cite this

Matsuda, T., & Strawderman, W. E. (2021). Predictive density estimation under the Wasserstein loss. Journal of Statistical Planning and Inference, 210, 53-63. https://doi.org/10.1016/j.jspi.2020.05.005

@article{46e0cd599e884dc088098b51df3a86c0,

title = "Predictive density estimation under the Wasserstein loss",

abstract = "We investigate predictive density estimation under the L2 Wasserstein loss for location families and location-scale families. We show that plug-in densities form a complete class and that the Bayesian predictive density is given by the plug-in density with the posterior mean of the location and scale parameters. We provide Bayesian predictive densities that dominate the best equivariant one in normal models. Simulation results are also presented.",

keywords = "Optimal transport, Predictive density, Shrinkage estimation, Wasserstein distance",

author = "Takeru Matsuda and Strawderman, {William E.}",

year = "2021",

month = jan,

doi = "10.1016/j.jspi.2020.05.005",

language = "English (US)",

volume = "210",

pages = "53--63",

journal = "Journal of Statistical Planning and Inference",

issn = "0378-3758",

publisher = "Elsevier",

}

TY - JOUR

T1 - Predictive density estimation under the Wasserstein loss

AU - Matsuda, Takeru

AU - Strawderman, William E.

PY - 2021/1

Y1 - 2021/1

N2 - We investigate predictive density estimation under the L2 Wasserstein loss for location families and location-scale families. We show that plug-in densities form a complete class and that the Bayesian predictive density is given by the plug-in density with the posterior mean of the location and scale parameters. We provide Bayesian predictive densities that dominate the best equivariant one in normal models. Simulation results are also presented.

AB - We investigate predictive density estimation under the L2 Wasserstein loss for location families and location-scale families. We show that plug-in densities form a complete class and that the Bayesian predictive density is given by the plug-in density with the posterior mean of the location and scale parameters. We provide Bayesian predictive densities that dominate the best equivariant one in normal models. Simulation results are also presented.

KW - Optimal transport

KW - Predictive density

KW - Shrinkage estimation

KW - Wasserstein distance

UR - http://www.scopus.com/inward/record.url?scp=85084995166&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85084995166&partnerID=8YFLogxK

U2 - 10.1016/j.jspi.2020.05.005

DO - 10.1016/j.jspi.2020.05.005

M3 - Article

AN - SCOPUS:85084995166

VL - 210

SP - 53

EP - 63

JO - Journal of Statistical Planning and Inference

JF - Journal of Statistical Planning and Inference

SN - 0378-3758

ER -