Entropy Matching Criterion and Statistical Properties of the Quantile-Based Multivariate Log-t Distribution with Application to Financial Data
Entropy · 06.11.2026 · 28 (11) · 1222
DOI: 10.3390/e28111222 ↗Kurz gefasst
Fachartikel in der Zeitschrift Entropy, online erschienen am 06.11.2026. Verfasst von Julián Andrés Angarita-Suárez und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Universidad de Antioquia.
Abstract (Originalsprache)
Positive, asymmetric and heavy-tailed random vectors are common in finance and the environmental sciences, where means may not exist. We study the quantile-based multivariate log-t (QLT) distribution, a reparameterization of the multivariate log-t (LT) distribution whose parameters are marginal quantiles at user-chosen levels and pairwise associations. It reduces to the LT family when all levels equal 1/2, converges to the quantile-based multivariate log-normal distribution as ν→∞, and has no finite mixed moment for any degrees of freedom ν. We derive its main properties and develop a population theory of the Entropy Matching Criterion (EMC), which matches the entropy of a standard multivariate t distribution with a k-nearest-neighbor entropy estimate of the standardized residuals. The criterion is calibrated at the true ν and identifies it locally but, in general, not globally, and its selection is displaced upward in simulations. Profile maximum likelihood is therefore preferable for estimating ν, and the EMC profile serves as a likelihood-free diagnostic of its identification; the EMC selection is preferable only for upper quantiles at the levels 0.90 and 0.95 under asymmetric data. An application to daily trading volumes illustrates the quantile interpretation and the model’s limitations.
Förderung
- Universidad de Antioquia
Zitieren
APA
Angarita-Suárez, J. A., Cataño-Salazar, D. H., & Morán-Vásquez, R. A. (2026). Entropy Matching Criterion and Statistical Properties of the Quantile-Based Multivariate Log-t Distribution with Application to Financial Data. Entropy, 28(11), 1222. https://doi.org/10.3390/e28111222
Vancouver
Angarita-Suárez JA, Cataño-Salazar DH, Morán-Vásquez RA. Entropy Matching Criterion and Statistical Properties of the Quantile-Based Multivariate Log-t Distribution with Application to Financial Data. Entropy. 2026;28(11):1222. doi:10.3390/e28111222
BibTeX
@article{AngaritaSurez2026,
author = {Angarita-Suárez, Julián Andrés and Cataño-Salazar, Duván Humberto and Morán-Vásquez, Raúl Alejandro},
title = {Entropy Matching Criterion and Statistical Properties of the Quantile-Based Multivariate Log-t Distribution with Application to Financial Data},
journal = {Entropy},
year = {2026},
volume = {28},
number = {11},
pages = {1222},
doi = {10.3390/e28111222}
}
EUVIMED