Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ LAReferencia - Red F...arrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
versions View all 1 versions
addClaim

Modelos lineales generalizados: regresión de rango reducido y reducción suficiente de dimensiones

Authors: Duarte, Sabrina Lorena;

Modelos lineales generalizados: regresión de rango reducido y reducción suficiente de dimensiones

Abstract

La respuesta a muchos de los problemas de interés en ciencias experimentales requieren el estudio de una o varias variables Y (respuesta/s) en función de otras variables X (predictores). Desde el punto de vista de la estadística, esto significa estudiar la distribución condicional de un vector Y r-dimensional, dado el vector X p-dimensional. Cuando el número de predictores p es grande, casi todos los métodos usados para estudiar esta relación incluye algún tipo de reducción en la dimensión de X. Los métodos estadísticos más recientes son aquellos establecidos bajo el paradigma de reducción suficiente de dimensiones. Su premisa es la obtención de una reducción de los predictores R(X) d-dimensional con d menor que p, sin que ésta pierda información acerca de la respuesta en el sentido que Y dado X tiene la misma distribución que Y dado R(X). En esta tesis trabajamos bajo el enfoque basado en la suposición de un modelo parámetrico para la regresión inversa X dado Y. El atractivo de este enfoque es que cuando la respuesta es univariada, X dado Y consta de p regresiones univariadas las cuales son sencillas de modelar, contrariamente a lo que ocurre cuando se modela Y dado X. El objetivo de esta tesis es desarrollar una metodología de reducción suficiente de dimensiones asumiendo que la distribución de X dado Y pertenece a una familia exponencial a k parámetros naturales . Para este modelo identificamos la reducción suficiente minimal, obtenemos estimadores de máxima verosimilitud para dicha reducción y estudiamos sus distribuciones asintóticas. Además, mostramos ejemplos y simulaciones para ilustrar las conexiones, diferencias y ventajas de nuestro método con los ya existentes. Para poder desarrollar este trabajo es necesario, en primer lugar, estudiar en detalle los modelos lineales generalizados multivariados de rango reducido.

The answer to many problems of interest in experimental sciences require the study of one or more variables Y (response/s) depending on other variables X (predictors). From the point of view of statistics, this means study the conditional distribution of a vector Y r-dimensional given the vector X p-dimensional. When the number of predictors p is large, almost all methods used to study this relationship includes some kind of reduction in the dimension of X. The most recent statistical methods are those established under the paradigm of sufficient dimension reduction. Its premise is to obtain a reduction in the predictors, R(X), d-dimensional with d less than p, without loss of information about the response in the sense that Y given X has the same distribution as Y given R (X). In this thesis we work under the approach based on the assumption of a parametric model for the inverse regression X given Y. The appeal of this approach is that when the answer is univariate, X given Y consists of p univariate regressions which are simple to model contrary to what happens when modeling Y given X. The aim of this thesis is to develop a methodology of sufficient dimension reduction assuming that the distribution of X given Y belongs to an exponential family with k natural parameters. For this model we identify sufficient minimal reduction, we obtain maximum likelihood estimators for the reduction and we study their asymptotic distributions. Moreover, we show examples and simulations to illustrate the connections, differences and advantages of our method with existing ones. To develop this work is necessary, first, to study in detail multivariate generalized linear models of rank reduced.

Fil: Duarte, Sabrina Lorena. Universidad Nacional del Litoral. Facultad de Ingeniería Química; Argentina.

Consejo Nacional de Investigaciones Científicas y Técnicas

Agencia Nacional de Promoción Científica y Tecnológica

Universidad Nacional del Litoral

Country
Argentina
Related Organizations
Keywords

Generalized linear models, Modelos lineales generalizados, Sufficient reduction, Familias exponenciales, Exponential family, Reduced rank regression, Reducción suficiente, Regresión de rango reducido

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Green