• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Doctoral Thesis
DOI
https://doi.org/10.11606/T.104.2024.tde-30072024-110832
Document
Author
Full name
Daniel Camilo Fuentes Guzman
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 2024
Supervisor
Committee
Louzada Neto, Francisco (President)
Diniz, Carlos Alberto Ribeiro
Gonzatto Junior, Oilson Alberto
Ramos, Pedro Luiz
Zeller, Camila Borelli
Title in English
Influence Diagnostics in Linear Censored Regression Models with Skew Scale Mixtures of Normal Distributions
Keywords in English
Algorithm
Censoring
Diagnostics
Distributions
EM
Influence
Linear
Mean-Shift
Mixtures of normal
Models
Outlier
Regression
Scale
Skew
SSMN
Abstract in English
In this research, we conducted studies on local and global influence diagnostics for Censored Linear Regression Models with Skew Scale Mixtures of Normal Distributions (SSMN-CR), proposed by Guzman, Ferreira and Zeller (2020). Initially, we discussed methods for generating censored data, specifically presenting methods to generate randomly censored data with both unilateral and interval censoring. Subsequently, we addressed case deletion and local influence diagnostics based on the Q function, inspired by the findings of Zhu et al. (2001) and Zhu and Lee (2001). To analyze the sensitivity of the maximum likelihood estimators of the SSMN-CR model parameters to small perturbations in assumptions and/or data, we considered various perturbation schemes, such as case weighting, explanatory variables, response variables, and perturbations in scale and skewness parameters. To illustrate the usefulness of the proposed methodology, we presented the analysis of a real dataset and three simulation studies.
Title in Portuguese
Diagnóstico de influência em modelos de regressão linear censurada com distribuições de misturas de escala assimétrica de normais
Keywords in Portuguese
Algoritmo
Assimetria
Censura
Diagnóstico
Distribuições
EM
Influência
Linear
Mean-Shift
Misturas de escala
Modelos
Normais
Outlier
Regressão
Abstract in Portuguese
Nesta pesquisa, conduzimos estudos de diagnóstico de influência local e global para modelos de regressão linear com censura e misturas de escala assimétrica de distribuições normais, propostos por Guzman, Ferreira and Zeller (2020) e denotados como SSMN-CR. Inicialmente, discutimos métodos para gerar dados censurados, apresentando especificamente métodos para gerar dados censurados aleatórios com censura unilateral e intervalar. Posteriormente, abordamos a exclusão de casos e o diagnóstico de influência local com base na função Q, inspirada nas descobertas de Zhu et al. (2001) e Zhu and Lee (2001). Para analisar a sensibilidade dos estimadores de máxima verossimilhança dos parâmetros do modelo SSMN-CR a pequenas perturbações nos pressupostos e/ou dados, consideramos vários esquemas de perturbação, como ponderação de casos, variáveis explicativas, variáveis resposta e perturbações nos parâmetros de escala e assimetria. Para ilustrar a utilidade da metodologia proposta, apresentamos a análise de um conjunto de dados reais e três estudos de simulação.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2024-07-30
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.