Diferenças

Aqui você vê as diferenças entre duas revisões dessa página.

--- projetos:gem2 [2011/07/02 17:22] – [Histórico] walmes
+++ projetos:gem2 [2011/10/18 10:07] (atual) – [section 5] eder
@@ Linha 16: / Linha 16: @@
     * [[pessoais:paulojus|Paulo Justiniano Ribeiro Jr]]
     * [[pessoais:silvia|Silvia Emiko Shimakura]]
+    * [[pessoais:elias|Elias Teixeira Krainski]]
     * [[pessoais:wbonat|Wagner Hugo Bonat]]
     * [[pessoais:walmes|Walmes Marques Zeviani]]
@@ Linha 25: / Linha 26: @@
 ----
-==== Scripts ====
+==== Scripts  & Códigos====
     * {{:pessoais:mixedeffects.r|Random Effects - Faraway}}
+    * {{:pessoais:reml.r|REML e ML}} Comparação entre verossimilhança (ML) e verossimilhança residual (REML) em modelos mistos
     * [[http://www.maths.bath.ac.uk/~jjf23/ELM/scripts/|Scritps]] Extending the Linear Model with R
     * [[http://www.maths.bath.ac.uk/~jjf23/ELM/mixchange.pdf| Random Effects]] Material Atualizado do livro
-    * [[http://www.ebah.com.br/content/ABAAAAG_sAC/modelos-lineares|João Gil de Luna ]]
+    * {{:pessoais:pql.r|PQL para dados Poisson com efeito aleatório individual não estruturado (Silvia)}}
 ----
+==== Artigos ====
+  *{{http://www.jstor.org/stable/pdfplus/2527669.pdf|The Estimation of Environmental and Genetic Trends from Records Subject to Culling}}
+  * {{http://www.sciencedirect.com/science/article/pii/S0022030278835905|Simulation of Examine Distributions of Estimators of Variances and Ratios of Variances}}
+  * {{http://www.jstor.org/stable/3001853|Estimation of Variance and Covariance Components}}
+  * {{http://www.sciencedirect.com/science/article/pii/S0022030291786013|C. R. Henderson: Contributions to Predicting Genetic Merit}}
+  * {{http://download.journals.elsevierhealth.com/pdfs/journals/0022-0302/PIIS002203027584776X.pdf|Rapid Method for Computing the Inverse of a Relationship Matrix}}
+  * {{http://www.jstor.org/pss/2529339|A simple method for computing the inverse of a numerator relationship matrix used in prediction of breeding values}}
+  * {{http://www.jstor.org/stable/2529430?&Search=yes&searchText=%22C.+R.+Henderson%22&list=hide&searchUri=%2Faction%2FdoBasicSearch%3FQuery%3Dau%253A%2522C.%2BR.%2BHenderson%2522%26wc%3Don&prevSearch=&item=3&ttl=373&returnArticleService=showFullText|Best Linear Unbiased Estimation and Prediction under a Selection Model}}
+  * {{http://www.jstor.org/stable/2530609?&Search=yes&searchText=%22C.+R.+Henderson%22&list=hide&searchUri=%2Faction%2FdoBasicSearch%3FQuery%3Dau%253A%2522C.%2BR.%2BHenderson%2522%26wc%3Don&prevSearch=&item=5&ttl=373&returnArticleService=showFullText|Variance-Covariance Matrix of Estimators of Variances in Unweighted Means ANOVA}}
+  * {{http://www.jstor.org/stable/2529264?seq=10|A Comparison of Variance Component Estimators}}
 ==== Slides, notas de aula e apostilas ====
@@ Linha 77: / Linha 91: @@
   * {{http://www.google.com/search?client=ubuntu&channel=fs&q=restricted+maximum+likelihood+estimation+of+variance+components+in+generalized+linear+mixed+models&ie=utf-8&oe=utf-8#hl=pt-BR&client=ubuntu&hs=tLi&channel=fs&q=restricted+maximum+likelihood+estimation+of+variance+components+in+generalized+linear+mixed+models+filetype:ps&aq=f&aqi=&aql=&oq=&pbx=1&bav=on.2,or.r_gc.r_pw.&fp=7992793a42cf0be6&biw=1760&bih=849|Thomas Kneib - Restricted maximum likelihood estimation of variance components in generalized linear mixed models}}; =)
   * {{http://research.utar.edu.my/CMS/ICMSA2010/ICMSA2010_Proceedings/files/poster/P-Rashidi.pdf|Fit generalized linear mixed models (GLMM), using of different likelihoods}};
+  * {{http://www.ebah.com.br/content/ABAAAAG_sAC/modelos-lineares|Modelos Lineares - João Gil de Luna}};
+  * {{http://nlp.stanford.edu/~manning/courses/ling289/GLMM.pdf|Generalized Linear Mixed Models (illustrated with R on Bresnan et al.’s datives data)}};
 ----
@@ Linha 109: / Linha 125: @@
 ==== Referências  ====
-=== Artigos ===
 === Livros ===
 <bibtex>
@@ Linha 186: / Linha 202: @@
 ----
-==== Códigos para modelos Mistos ====
-=== Exemplo Random Effects Faraway  ===
-<code R>
-### Modelo mistos
-### Exemplo retirado do capitulo Random Effeets do Faraway
-require(faraway)
-require(lme4)
-###------------------------------------------------------------###
-### Modelo simples com apenas um fator
-data(pulp)
-str(pulp)
-summary(pulp)
-op <- options(contrasts=c("contr.sum", "contr.poly"))
-### Efeitos Fixos
-lmod <- aov(bright ~ operator, pulp)
-summary(lmod)
-summary.lm(lmod)
-coef(lmod)
-sigma2alfa <- (0.447-0.106)/5
-sigma2alfa
-model.matrix(lmod)
-modelo <- matrix(apply(model.matrix(lmod),1,function(x){coef(lmod)*x}),ncol=4,byrow=TRUE)
-compoF <- cbind(pulp$bright,modelo,lmod$res)
-colnames(compoF) <- c('var','Intercept','operator1','operator2','operator3','erro')
-compoF
-rowSums(compoF[,-1])
-compoF[,1]
-summary(lmod)
-SQV <- sum(compoF[,1]^2)
-SQI <- sum(compoF[,2]^2)
-SQT <-sum(rowSums(compoF[,3:5])^2)
-SQE <- sum(compoF[,6]^2)
-SQTc <- SQV-SQI
-SQTc
-SQT
-SQE
-summary(lmod)
-SQTc
-SQT+SQE
-### Efeitos aleatorios
-mmod <- lmer(bright ~ 1+(1|operator), pulp)
-summary(mmod)
-str(mmod)
-compoM <- matrix(0,nrow=20,ncol=3)
-compoM[,1] <- fixef(mmod)
-compoM[,2] <- rep(ranef(mmod)$operator[,1],each=5)
-compoM[,3] <- residuals(mmod)
-compoM <- cbind(pulp$bright,compoM)
-colnames(compoM) <- c('var','intercepto','efeito','erro')
-compoM
-smod <- lmer(bright ~ 1+(1|operator), pulp,method="ML")
-summary(smod)
-compoS <- matrix(0,nrow=20,ncol=3)
-compoS[,1] <- fixef(smod)
-compoS[,2] <- rep(ranef(smod)$operator[,1],each=5)
-compoS[,3] <- residuals(smod)
-compoS <- cbind(pulp$bright,compoS)
-colnames(compoS) <- c('var','intercepto','efeito','erro')
-compoS
-SQE <- t(compoS[,'erro'])%*%compoS[,'erro']
-SQE
-rowSums(compoS[,-1])
-pulp$bright
-## Inferencia
-# Modelo nulo
-nullmod <- lm (bright ~ 1, pulp)
-summary(nullmod)
-##LRT
-tLRT <- as.numeric(2*(logLik(smod)-logLik(nullmod)))
-tLRT
-pchisq(tLRT,1,lower=FALSE)
-## Comparando os dois modelos
-anova(mmod,smod)
-#bootstrap
-y <- simulate(nullmod,50)
-#y
-### Visualizando a simulação
-hist(pulp$bright,prob=T)
-apply(y,2,function(x){lines(density(x))})
-lines(density(pulp$bright),lw=2,col='red')
-## Bootstrap
-lrstat <- numeric(1000)
-for(i in 1:1000){
-  y <- unlist(simulate(nullmod))
-  bnull <- lm(y ~ 1)
-  balt <- lmer(y~1 + (1|operator),pulp,method="ML")
-  lrstat[i] <- as.numeric(2*(logLik(balt)-logLik(bnull)))
-}
-hist(lrstat)
-summary(lrstat)
-mean(lrstat < 0.0001)
-# pvalor bootstrap
-pb <- mean(lrstat > 2.5684)
-pb
-## Erro padrão boosstrap
-standErro <- sqrt(pb*(1-pb)/1000)
-standErro
-# Prediçao
-ranef(mmod)$operator
-cc <- model.tables(lmod)
-cc
-###Shirikange
-shir <- cc[[1]]$operator/ranef(mmod)$operator
-shir
-blups <- fixef(mmod)+ranef(mmod)$operator
-blups
-## Residuos
-par(mfrow=c(1,2))
-qqnorm(resid(mmod),main="")
-plot(fitted(mmod),resid(mmod),xlab="Fitted",ylab="Residuals")
-abline(0,0)
-###------------------------------------------------------------###
-### Modelo com efeito de bloco
-rm(list=ls())
-data(penicillin)
-summary(penicillin)
-## Modelo fixo
-op <- options(contrasts=c("contr.sum", "contr.poly"))
-lmod <- aov(yield ~ blend + treat, penicillin)
-summary(lmod)
-coef(lmod)
-compoF <- matrix(0,nrow=20,ncol=4)
-compoF[,1] <- coef(lmod)[1]
-compoF[,2] <- rep(c(coef(lmod)[2:5],0-sum(coef(lmod)[2:5])),each=4)
-compoF[,3] <- rep(c(coef(lmod)[6:8],0-sum(coef(lmod)[6:8])),5)
-compoF[,4] <- lmod$res
-compoF <- cbind(penicillin$yield,compoF)
-colnames(compoF) <- c('var','intercepto','bloco','efeito','erro')
-compoF
-rowSums(compoF[,-1])
-compoF[,1]
-## Modelo misto
-mmod <- lmer (yield ~ treat + (1|blend), penicillin)
-summary(mmod)
-ranef(mmod)$blend
-fixef(mmod)
-compoM <- matrix(0,nrow=20,ncol=4)
-compoM[,1] <- fixef(mmod)[1]
-compoM[,2] <- rep(ranef(mmod)$blend[,1],each=4)
-compoM[,3] <- rep(c(fixef(mmod)[2:4],0-sum(fixef(mmod)[2:4])),5)
-compoM[,4] <- residuals(mmod)
-compoM <- cbind(penicillin$yield,compoM)
-colnames(compoM) <- c('var','intercepto','bloco','efeito','erro')
-compoM
-rowSums(compoM[,-1])
-compoM[,1]
-anova(mmod)
-amod <- lmer (yield ~ treat + (1|blend), penicillin,method="ML")
-nmod <- lmer (yield ~ 1 + (1|blend), penicillin,method="ML")
-anova(amod,nmod)
-## bootstrap
-lrstat <- numeric(1000)
-for(i in 1:1000){
-  ryield <- unlist(simulate(nmod))
-  nmodr <- lmer(ryield ~ 1 + (1|blend), penicillin,method="ML")
-  amodr <- lmer(ryield ~ treat + (1|blend), penicillin,method="ML")
-  lrstat[i] <- 2*(logLik(amodr)-logLik(nmodr))
-}
-plot(qchisq((1:1000)/1001,3),sort(lrstat),xlab=expression(chi[3]^2),ylab="Simulated LRT")
-abline(0,1)
-mean(lrstat > 4.05)
-rmod <- lmer(yield ~ treat + (1|blend), penicillin)
-nlmod <- lm(yield ~ treat, penicillin)
-* (logLik(rmod)-logLik(nlmod,REML=TRUE))
-lrstatf <- numeric(1000)
-for(i in 1:1000){
-  ryield <- unlist(simulate(nlmod))
-  nlmodr <- lm(ryield ~ treat, penicillin)
-  rmodr <- lmer(ryield ~ treat + (1|blend), penicillin)
-  lrstatf [i] <- 2*(logLik(rmodr)???logLik(nlmodr,REML=TRUE))
-  }
-mean(lrstatf < 0.00001)
-cs <- lrstatf[lrstatf > 0.00001]
-ncs <- length(cs)
-plot(qchisq((1:ncs)/(ncs+1),1),sort(cs),xlab=expression(chi[1]^2),ylab="Simulated LRT")
-abline (0,1)
-mean(lrstatf > 2.7629)
-###------------------------------------------------------------###
-### Modelo em parcelas Subdivididas
-data(irrigation)
-str(irrigation)
-summary(irrigation)
-attach(irrigation)
-### Modelo com mais parametros que variáveis
-#lmod <- lmer (yield ~ irrigation * variety + (1|field) + (1|field:variety),data=irrigation)
-lmodr <- lmer (yield ~ irrigation * variety + (1|field),data=irrigation)
-logLik(lmodr)
-summary(lmodr)
-anova(lmodr)
-mod <- aov(yield ~ irrigation * variety + Error(field),data=irrigation)
-summary(mod)
-mod <- lm(yield ~ irrigation * variety+field/variety,data=irrigation)
-anova(mod)
-model.matrix(mod)
-</code>
-----

Diferenças

Navegação

Busca

Ferramentas

QR Code