• 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
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.11.2007.tde-16102007-113618
Document
Author
Full name
Pamela Piovesan
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
Piracicaba, 2007
Supervisor
Committee
Dias, Carlos Tadeu dos Santos (President)
Govone, José Silvio
Piedade, Sonia Maria de Stefano
Title in Portuguese
Validação cruzada com correção de autovalores e regressão isotônica nos modelos AMMI
Keywords in Portuguese
Análise de dados
Análise multivariada
Estatística aplicada
Genótipos
Meio ambiente
Abstract in Portuguese
Neste trabalho apresenta-se a aplicação dos modelos AMMI para um estudo detalhado do efeito das interações entre genótipos e ambientes em experimentos multiambientais. Através da decomposição da soma de quadrados dessas interações, busca-se selecionar o número de termos que explicam essa interação, descartando o ruído presente na mesma. Há duas maneiras para a escolha desses termos: validação cruzada e teste de hipóteses. O foco será na validação cruzada pela vantagem de ser um critério "preditivo" de avaliação. São apresentados dois métodos de validação cruzada, métodos esses esboçados por Eastment e Krzanowski (1982) e Gabriel (2002). Esses métodos utilizam a decomposição por valores singulares para obter os autovalores referentes à matriz de interações, cuja soma de quadrados nos dá exatamente a soma de quadrados da interação. Como esses autovalores são superestimados ou subestimados (ARAÚJO; DIAS, 2002), essas técnicas de validação serão aperfeiçoadas através da correção desses autovalores e, para reordená-los, será utilizada a regressão isotônica. Será realizado um estudo comparativo entre esses métodos, através de dados reais.
Title in English
Cross-validation with eigenvalue correction and isotonic regression in models AMMI
Keywords in English
Applied statistic
Data analysis
Environment
Genotype
Multivariate analysis
Abstract in English
This paper presents the application of AMMI models for a thorough study about the effect of the interaction between genotypes and environments in multi-environments experiments. Through the decomposition of the sum of squares of these interactions, one searches to select the number of terms that explains this interaction, discarding its noise in. There are two ways for choosing these terms: cross-validation and hypotheses test. The focus will be on the crossvalidation for its advantage of being one prediction criterion of evaluation. Two methods of cross-validation are presented , both outlined by Eastment and Krzanowski (1982) and Gabriel (2002). These methods use the decomposition by singular values in order to obtain eigenvalues referred to the matrix of interactions, whose sum of squares accurately gives us the sum of squares of the interation. As these eigenvalues either over- or underestimated (ARAÚJO; DIAS, 2002), these techniques of validation will be improved through the correction of these eigenvalues and, in order to rearrange them, isotonic regression will be used . A comparative study between these methods through real data will be carried out.
 
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.
PamelaPiovesan.pdf (342.29 Kbytes)
Publishing Date
2007-10-24
 
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.