INFORMATION CRITERIA AND STOCHASTIC COMPLEXITY

Mostrar todas las versiones(2)

The objective criteria for the selection of the order of an autoregressive model can be classified into non-Bayesians, which are those based on the minimization of the prediction error and on the information measures, and Bayesians. The former group assume the validity of the hypothesis that every p...

Descripción completa

Detalles Bibliográficos
Autores principales: Gonzalez, Mirta, Landro, Alberto H.
Formato: Artículo publishedVersion
Lenguaje:Español
Publicado: Centro de Investigación en Métodos Cuantitativos Aplicados a la Economía y la Gestión (CMA) 2018
Acceso en línea:https://ojs.economicas.uba.ar/RIMF/article/view/1415
https://repositoriouba.sisbi.uba.ar/gsdl/cgi-bin/library.cgi?a=d&c=modelfin&d=1415_oai
Aporte de:
Repositorio Digital de la Universidad de Buenos Aires (UBA) de Universidad de Buenos Aires
id I28-R145-1415_oai
record_format dspace
spelling I28-R145-1415_oai2025-02-11 Gonzalez, Mirta Landro, Alberto H. 2018-07-30 The objective criteria for the selection of the order of an autoregressive model can be classified into non-Bayesians, which are those based on the minimization of the prediction error and on the information measures, and Bayesians. The former group assume the validity of the hypothesis that every process is affected by its infinite past and provides asymptotically efficient estimators, while the Bayesians rely on the denial of the Church-Turing thesis and provide consistent estimators. To avoid the disjunctive generated by this classification, it is proposed to characterize the model through the definition of stochastic complexity. The application of this concept and the postulates of the convergence theorems of the complexity measures allow in addition to demonstrate the optimal condition of the penalty term of the Schwarz selection criterion. Los criterios objetivos de selección del orden de un modelo autorregresivo pueden ser clasificados en no-Bayesianos -basados en la minimización del error de predicción y las medidas de información- y Bayesianos. La diferencia entre ambos radica en que los primeros asumen como punto de partida la validez de la hipótesis de que todo proceso está afectado por su infinito pasado y proporcionan estimadores asintóticamente eficientes en tanto que los Bayesianos se basan en la negación de la tesis de Church-Turing y proporcionan estimadores consistentes. A fin de evitar la disyuntiva que genera esta clasificación, en este trabajo se propone caracterizar al modelo utilizando la definición de complejidad estocástica. La aplicación de este concepto y los postulados de los teoremas de convergencia de las medidas de complejidad permiten demostrar, además, la condición de óptimo del término de penalización del criterio de selección de Schwarz. application/pdf https://ojs.economicas.uba.ar/RIMF/article/view/1415 spa Centro de Investigación en Métodos Cuantitativos Aplicados a la Economía y la Gestión (CMA) https://ojs.economicas.uba.ar/RIMF/article/view/1415/2034 Revista de Investigación en Modelos Financieros; Vol. 1 (2018): Revista de Investigación en Modelos Financieros; 21-40 2250-6861 2250-687X INFORMATION CRITERIA AND STOCHASTIC COMPLEXITY CRITERIOS DE INFORMACIÓN Y COMPLEJIDAD ESTOCÁSTICA info:eu-repo/semantics/article info:eu-repo/semantics/publishedVersion https://repositoriouba.sisbi.uba.ar/gsdl/cgi-bin/library.cgi?a=d&c=modelfin&d=1415_oai
institution Universidad de Buenos Aires
institution_str I-28
repository_str R-145
collection Repositorio Digital de la Universidad de Buenos Aires (UBA)
language Español
orig_language_str_mv spa
description The objective criteria for the selection of the order of an autoregressive model can be classified into non-Bayesians, which are those based on the minimization of the prediction error and on the information measures, and Bayesians. The former group assume the validity of the hypothesis that every process is affected by its infinite past and provides asymptotically efficient estimators, while the Bayesians rely on the denial of the Church-Turing thesis and provide consistent estimators. To avoid the disjunctive generated by this classification, it is proposed to characterize the model through the definition of stochastic complexity. The application of this concept and the postulates of the convergence theorems of the complexity measures allow in addition to demonstrate the optimal condition of the penalty term of the Schwarz selection criterion.
format Artículo
publishedVersion
author Gonzalez, Mirta
Landro, Alberto H.
spellingShingle Gonzalez, Mirta
Landro, Alberto H.
INFORMATION CRITERIA AND STOCHASTIC COMPLEXITY
author_facet Gonzalez, Mirta
Landro, Alberto H.
author_sort Gonzalez, Mirta
title INFORMATION CRITERIA AND STOCHASTIC COMPLEXITY
title_short INFORMATION CRITERIA AND STOCHASTIC COMPLEXITY
title_full INFORMATION CRITERIA AND STOCHASTIC COMPLEXITY
title_fullStr INFORMATION CRITERIA AND STOCHASTIC COMPLEXITY
title_full_unstemmed INFORMATION CRITERIA AND STOCHASTIC COMPLEXITY
title_sort information criteria and stochastic complexity
publisher Centro de Investigación en Métodos Cuantitativos Aplicados a la Economía y la Gestión (CMA)
publishDate 2018
url https://ojs.economicas.uba.ar/RIMF/article/view/1415
https://repositoriouba.sisbi.uba.ar/gsdl/cgi-bin/library.cgi?a=d&c=modelfin&d=1415_oai
work_keys_str_mv AT gonzalezmirta informationcriteriaandstochasticcomplexity
AT landroalbertoh informationcriteriaandstochasticcomplexity
AT gonzalezmirta criteriosdeinformacionycomplejidadestocastica
AT landroalbertoh criteriosdeinformacionycomplejidadestocastica
_version_ 1825551174860800000

Ejemplares similares