Trabalhos Aprovados

 

Atenção: As comunicações foram provisoriamente aceitas para apresentação no 18o SINAPE. Apenas os trabalhos em que pelo menos um dos autores tiver, até 16/06, pago o boleto de inscrição do congresso terão a aceitação definitiva, os demais não poderão ser apresentados no evento. Verifique se o seu trabalho satisfaz esse requisito.

 

Filtrar por categoria/formato/palavra:


 
Formato Título
Pôster
A DISTRIBUIÇÃO BETA BINOMIAL NEGATIVA
Probabilidade e Processos Estocásticos
Autor 1: Ana Paula Coelho Madeira (UFLA)
anapaulacoelho21@yahoo.com.br
Autor 2: Lucas Monteiro Chaves (UFLA)
lucas@ufla.br
Autor 3: Maria do Carmo Toledo Costa (UFLA)
mctoledo@ufla.br
Abstract:
A mistura da distribuição Binomial Negativa com a distribuição Beta é obtida e denominada distribuição Beta-Binomial Negativa. Como aplicação é realizado um ajuste a dados do número de plantas Salicornia stricta, analisados anteriormente por Bliss-Fisher.
Pôster
AVALIAÇÃO DE ACURÁCIA E PRECISÃO DE MÉTODOS DE ESTIMAÇÃO DO PARÂMETRO REGRESSIVO DE UM MODELO DE ANÁLISE DE COVARIÂNCIA COM ERRO DE MEDIDA SOBRE DIFERENTES DISTRIBUIÇÕES
Estatística em Agronomia e Biologia
Autor 1: Tiago Almeida Oliveira (Esalq/USP)
tadolive@esalq.usp.br
Autor 2: Augusto Ramalho Morais (UFLA)
armorais@ufla.br
Autor 3: Marcelo Angelo Cirillo (UFLA)
macufla@gmail.com
Autor 4: Marcelino Alves Pascoa (Esalq/USP)
mapascoa@esalq.usp.br
Autor 5: Renato Nunes Pereira (Esalq/USP)
rpereira@esalq.usp.br
Abstract:
The present work approaches the covariance analysis model with one factor and measurement error in the covariate. Accuracy and precision of two estimators suggested in the literature were evaluated through data simulation, for estimating parameters of a regression model with measurement error. So called ``Plug-in'' method estimates the real value based on the observed ones and then uses the common function for estimating the desired parameter. The other estimator, known as bias smoother, only performs a bias correction on the usual estimator by computing a factor. Behavior of both estimators was studied under different residual distributions, goodness of fit and sample sizes. Results suggest that the ``Plug-in'' estimator presented the best performance both for accuracy and precision under normality, for the distinct evaluated situations. It is worth noting that, in covariance analysis model, the high the sample size, the better for accuracy and precision. When the estimators had been evaluated in the model of ANCOVA with the residues distri\-buted for Gamma, the same ones had gotten the worse performance in relation when they were evaluated by the others distributions.
Oral
A abordagem de Cadastro Duplo (Dual Frame): Estimação assistida por modelos lineares com aplicação em pesquisas agropecuárias
Concurso de Dissertação de Mestrado
Aluno: HEMÍLIO COÊLHO (UFPE)
hemilio@gmail.com
Orientador: CRISTIANO FERRAZ (UFPE)
cferraz@de.ufpe.br
Abstract:
O presente trabalho apresenta técnicas de estimação de totais e médias populacionais sob a abordagem de cadastro duplo (dual frame), incorporando informações de variáveis auxiliares ao processo de estimação, sob a assistência de modelos lineares. A abordagem de cadastro duplo é definida como um levantamento amostral onde dois cadastros são utilizados para identificar elementos de uma única população-alvo. Considerando este contexto, diversos estimadores assistidos por modelos lineares são propostos. Seus desempenhos são comparados com estimadores encontrados na literatura e avaliados com base no método de simulação de Monte Carlo. Resultados são apresentados para o plano de amostragem aleatória simples e para o plano de amostragem estratificada. As técnicas propostas têm potencialidade de aplicação direta em pesquisas agropecuárias.
Oral
A Bayesian Approach for Temporal Clustering
Métodos Bayesianos
Autor 1: João Vitor Dias Monteiro (UFMG)
piocitos@gmail.com
Autor 2: Rosângela Helena Loschi (UFMG)
loschi@est.ufmg.br
Autor 3: Renato Martins Assunção (UFMG)
assuncao@est.ufmg.br
Abstract:
Many subject areas are share interest in temporal clusters identification problems. A temporal cluster is an aggregation of subsequential times such that the counts are likely to come from a simple stochastic model. In time series, Bayesian partition models aim at partitioning the entire observation period into disjoint temporal clusters. We developed a hierarchical Bayesian model to find out those temporal clusters in a data sequence when the observations have Poisson distribution conditional on their unobserved means. In this paper, we present our proposed model and its connections with the Product Partition Model (PPM) introduced by Hartigan (1990). Also we present preliminary results using a famous series of coal mine accidents in England.
Pôster
A class of asymptotically normal degenerate quasi U-statistics
Inferência Estatística
Autor 1: Pranab Kumar Sen (UNC-Chapel Hill)
pksen@bios.unc.edu
Autor 2: Aluísio Pinheiro (Unicamp)
pinheiro@ime.unicamp.br
Autor 3: Hildete Prisco Pinheiro (Unicamp)
hildete@ime.unicamp.br
Abstract:
Some quasi U-statistics, unlike other variants of U-statistics, arising in distance based tests for homogeneity of groups, have first-order stationary kernels of degree 2, and yet they enjoy asymptotic normality under suitable hypotheses of invariance. Central limit theorems for a more general class of quasi U-statistics with possibly higher-order stationarity (and degree) are formulated with the aid of appropriate martingale (array) characterizations as well as permutational invariance structures. AMS : Primary - 60F05, 62F12; Secondary - 62G20, 92D20. Keywords: Genomics, Hamming distance, Martingale, Orthogonal system, Permutation measure, Second-order asymptotics, Higher-order decomposability.
Pôster
A classification of binary asymmetric regression models: The use of BRMUW in an application to the decision to erradicate illegal crops of coca leaf
Econometria, Atuária e Finanças
Autor 1: Jorge Luis Bazan (PUCP)
jlbazan@pucp.edu.pe
Autor 2: Oscar Edgardo Millones (PUCP)
omillones@pucp.edu.pe
Abstract:
Abstract The binary regression model, which is based on the logit symmetric link, has been traditionally used in classical econometric literature. The purpose of this paper is to introduce binary regression models with rather asymmetric links, in cases where this asymmetry is more appropriate to the researcher. The paper includes an utility program called BRMUW (Bayesian Regression Model using WinBugs) which facilitates the writing of the syntax that is required to implement the models so reviewed. It also generates the syntax for the structure of the data being used. The method is illustrated with data from a sample of farmers who consider the decision to eradicate illegal crops of coca leaf. Factors associated with this decision are also explored. Keywords: asymmetric links, binary regression, bayesian inference, discrete choice econometric models, WinBUGS.
Oral
A Continuous Non Parametric Lifetime Value Estimator
Estatística em Ciências Sociais Aplicadas (Administração, Economia, Sociologia, Psicologia, etc.)
Autor 1: Osvaldo Anacleto Jr (Banco Itaú)
osvaldo.anacleto@gmail.com
Autor 2: Francisco Louzada-Neto (DEs-UFSCar)
dfln@power.ufscar.br
Abstract:
Many companies are interested in an estimation of their customers revenue. Lifetime value, an approximate measure of the profit earned with each customer in a time period, is a widely used approach to tackle this problem. In this work, we present a statistical procedure to determine the lifetime value based on survival analysis techniques. We apply the methodology in a numerical example.
Pôster
A cópula limite associada as k maiores estatísticas de ordem para amostras i.i.d
Probabilidade e Processos Estocásticos
Autor 1: Marco Aurélio Sanfins (UFRJ)
marcoa@dme.ufrj.br
Abstract:
Abstract In this work we obtain the expression of the Bi-extremal copula, the copula pertaining to the limit distribution of the two largest order statistics of iid samples. Properties of the Bi-extremal copula are derived, including the Kendall's coeficient and the tail dependence coeficient. Theoretical and simulated examples are worked out. As applications we propose the construction of meta-Bi-extremal distributions and provide an illustration of a modeling situation focusing in extreme risks. Also we obtain the expression of the K-extremal copula.
Pôster
A distribuição t-assimétrica univariada: propriedades e inferência
Concurso de Dissertação de Mestrado
Aluno: Luciana Godoi (USP )
graziela@ime.usp.br
Orientador: Márcia D'Elia Branco (USP)
mbranco@ime.usp.br
Abstract:
Nesta dissertação são apresentados diferentes resultados em torno da distribuição t-assimétrica univariada, a qual inclui como casos particulares as distribuições t-Student, normal-assimétrica e normal. Inicialmente, propõem-se várias maneiras de caracterizar esta distribuição e algumas propriedades são apresentadas. Questões inferenciais são discutidas sob as perspectivas clássica e bayesiana. Sob a primeira, consideramos os métodos dos momentos e de máxima verossimilhança e apresentamos possíveis modelos gráficos de diagnóstico. Através de um estudo de simulação, avaliamos o desempenho destes dois tipos de estimadores e concluímos que, em geral, amostras com grandes assimetrias apresentam estimativas de máxima verossimilhança mais próximas do verdadeiro valor do parâmetro do que aquelas obtidas via método dos momentos. No entanto, observamos um grande viés em ambas as estimativas. Os estimadores clássicos apresentam também problemas teóricos. Sob a perspectiva bayesiana, discutimos possíveis especificações a priori para os parâmetros da distribuição t-assimétrica. A fim de facilitar a implementação computacional do método bayesiano, apresentamos diferentes formas hierárquicas de representar o modelo t-assimétrico. A metodologia bayesiana mostrou-se mais eficiente para estimar os parâmetros de assimetria e graus de liberdade. Por fim, as inferências propostas anteriormente são aplicadas em dois conjuntos de dados: notas dos alunos do curso de MAE-0116 e uma amostra simulada.
Pôster
A distribui¸c˜ao Weibull Geométrica
Inferência Estatística
Autor 1: Wagner Barreto Souza (UFPE)
wagnerbs85@hotmail.com
Autor 2: Alice Lemos de Morais (UFPE)
alice.lm@hotmail.com
Abstract:
Neste trabalho é introduzida a distribuição Weibull geométrica (WG). Esta distribuição generaliza a distribuição exponencial geométrica (EG) [Adamidis e Loukas, 1998]. A função de risco da distribui¸c˜ao EG ´e estritamente decrescente, enquanto a fun¸c˜ao de risco da WG pode assumir mais formas. Calculamos as fun¸c˜oes de distribui¸c˜ao e risco da distribui¸c˜ao WG, al´em da densidade da i-´esima estat´ıstica de ordem. Apresentamos express˜oes para os momentos da distribui¸c˜ao WG e para os momentos da i-´esima estat´ıstica de ordem. A estima¸c˜ao por m´axima verossimilhan¸ca ´e discutida e um algoritmo EM (Dempster et al., 1977; McLachlan e Krishnan, 1997) ´e apresentado para estima¸c˜ao dos parˆametros. Al´em disso, obtemos a matriz de informa¸c˜ao de Fisher e discutimos inferˆencia no modelo WG. Por fim, aplica¸c˜oes a dados reais ´e apresentada.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54