A unified method for generating closed-form point estimators for exponential families: An example with the beta distribution applied to proportions of land used for farming

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Vila, Roberto, Saulo, Helton
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913993879191552
author Vila, Roberto
Saulo, Helton
author_facet Vila, Roberto
Saulo, Helton
contents We show that, after a simple power-transform reparameterization of the (vector) exponential family, the solutions to the likelihood equations coincide with moment-type estimating equations. This equivalence enables a unified route to closed-form point estimators for multi-parameter models that typically lack explicit maximum likelihood (ML) solutions. Within this framework we (i) recover, as special cases, several recent closed-form estimators from the literature; (ii) derive new families of estimators indexed by monotone transformations $g$; and (iii) establish strong consistency and asymptotic normality under mild regularity, including a dominated differentiation condition. As a detailed illustration, we derive closed-form estimators for parameters that index the beta distribution. A Monte Carlo simulation study is carried out to evaluate and compare the performance of proposed and existing estimators. Finally, we illustrate the approach with a novel municipal data set regarding proportions of land used for farming in Roraima (Brazil), which, to the best of our knowledge, has not been analyzed in the literature before, demonstrating the method's practical usefulness.
format Preprint
id arxiv_https___arxiv_org_abs_2508_12169
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A unified method for generating closed-form point estimators for exponential families: An example with the beta distribution applied to proportions of land used for farming
Vila, Roberto
Saulo, Helton
Methodology
We show that, after a simple power-transform reparameterization of the (vector) exponential family, the solutions to the likelihood equations coincide with moment-type estimating equations. This equivalence enables a unified route to closed-form point estimators for multi-parameter models that typically lack explicit maximum likelihood (ML) solutions. Within this framework we (i) recover, as special cases, several recent closed-form estimators from the literature; (ii) derive new families of estimators indexed by monotone transformations $g$; and (iii) establish strong consistency and asymptotic normality under mild regularity, including a dominated differentiation condition. As a detailed illustration, we derive closed-form estimators for parameters that index the beta distribution. A Monte Carlo simulation study is carried out to evaluate and compare the performance of proposed and existing estimators. Finally, we illustrate the approach with a novel municipal data set regarding proportions of land used for farming in Roraima (Brazil), which, to the best of our knowledge, has not been analyzed in the literature before, demonstrating the method's practical usefulness.
title A unified method for generating closed-form point estimators for exponential families: An example with the beta distribution applied to proportions of land used for farming
topic Methodology
url https://arxiv.org/abs/2508.12169