1 |
Problems in generalized linear model selection and predictive evaluation for binary outcomesTen Eyck, Patrick 15 December 2015 (has links)
This manuscript consists of three papers which formulate novel generalized linear model methodologies.
In Chapter 1, we introduce a variant of the traditional concordance statistic that is associated with logistic regression. This adjusted c − statistic as we call it utilizes the differences in predicted probabilities as weights for each event/non- event observation pair. We highlight an extensive comparison of the adjusted and traditional c-statistics using simulations and apply these measures in a modeling application.
In Chapter 2, we feature the development and investigation of three model selection criteria based on cross-validatory c-statistics: Model Misspecification Pre- diction Error, Fitting Sample Prediction Error, and Sum of Prediction Errors. We examine the properties of the corresponding selection criteria based on the cross- validatory analogues of the traditional and adjusted c-statistics via simulation and illustrate these criteria in a modeling application.
In Chapter 3, we propose and investigate an alternate approach to pseudo- likelihood model selection in the generalized linear mixed model framework. After outlining the problem with the pseudo-likelihood model selection criteria found using the natural approach to generalized linear mixed modeling, we feature an alternate approach, implemented using a SAS macro, that obtains and applies the pseudo-data from the full model for fitting all candidate models. We justify the propriety of the resulting pseudo-likelihood selection criteria using simulations and implement this new method in a modeling application.
|
2 |
penalized: A MATLAB toolbox for fitting generalized linear models with penaltiesMcIlhagga, William H. 07 August 2015 (has links)
Yes / penalized is a
exible, extensible, and e cient MATLAB toolbox for penalized maximum
likelihood. penalized allows you to t a generalized linear model (gaussian, logistic,
poisson, or multinomial) using any of ten provided penalties, or none. The toolbox can
be extended by creating new maximum likelihood models or new penalties. The toolbox
also includes routines for cross-validation and plotting.
|
3 |
Models for target detection times.Bae, Deok Hwan January 1989 (has links)
Approved for public release; distribution in unlimited. / Some battlefield models have a component in them which models the time it takes for an observer to detect a target. Different observers may have different mean detection times due to various factors such as the type of sensor used, environmental conditions, fatigue of the observer, etc. Two parametric models for the distribution of time to target detection are considered which can incorporate these factors. Maximum likelihood estimation procedures for the parameters are described. Results of simulation experiments to study the small sample behavior of the estimators are presented. / http://archive.org/details/modelsfortargetd00baed / Major, Korean Air Force
|
4 |
Modelos paramétricos para séries temporais de contagem / Parametric models for count time seriesMilhorança, Igor André 14 May 2014 (has links)
Diversas situações práticas exigem a análise de series temporais de contagem, que podem apresentar tendência, sazonalidade e efeitos de variáveis explicativas. A motivação do nosso trabalho é a análise de internações diárias por doenças respiratórias para pessoas com mais que 65 anos residentes no município de São Paulo. O efeito de variáveis climáticas e concentrações de poluentes foram incluídos nos modelos e foram usadas as funções seno e cosseno com periodicidade de um ano para explicar o padrão sazonal e obter os efeitos das variáveis climáticas e poluentes controlando essa sazonalidade. Outro aspecto a ser considerado é a inclusão da população nas análises de modo que a interpretação dos efeitos seja para as taxas diárias de internações. Diferentes modelos paramétricos foram propostos para as internações. O mais simples é o modelo de regressão linear para o logaritmo das taxas. Foram ajustados os modelos lineares generalizados (MLG) para as internações com função de ligação logaritmo e com a população como offset, por este modelo permitir o uso das distribuições Poisson e Binomial Negativa, usadas para dados de contagem. Devido à heteroscedasticidade extra, foram propostos modelos GAMLSS incluindo variáveis para explicar o desvio padrão. Foram ajustados modelos ARMA e GARMA, por incluírem uma estrutura de correlação serial. O objetivo desse trabalho é comparar as estimativas, os erros padrões, a cobertura dos intervalos de confiança e o erro quadrático médio para o valor predito segundo os vários modelos e a escolha do modelo mais apropriado, que depende da completa análise de resíduos, geralmente omitida na literatura. O modelo GARMA com distribuição Binomial Negativa apresentou melhor ajuste, pois os erros parecem seguir a distribuição proposta e tem baixa autocorrelação, além de ter tido uma boa cobertura pelo intervalo de confiança e um baixo erro quadrático médio. Também foi analisado o efeito da autocorrelação dos dados nas estimativas nos vários modelos baseado em dados simulados. / Many practical situations require the analysis of time series of counts, which may present trend, seasonality and effects of covariates. The motivation of this work is the analysis of daily hospital admissions for respiratory diseases in people over 65 living in the city of São Paulo. The effect of climatic variables and concentrations of pollutants were included in the models and the sine and cosine functions with annual period were included to explain the seasonal pattern and obtain the effects of pollutants and climatic variables partially controlled by this seasonality. Another aspect to be considered is the inclusion of the population in the analys es in order to interpret the effects based on daily hospitalization rates . Different parametric models have been proposed for hospitalizations. The simplest is the linear regression model for the logarithm of the hospitalization rate. The generalized linear models (GLM) were adjusted for daily admissions with logarithmic link function and the population as offset to consider the Poisson and Negative Binomial distributions for counting data. Due to the extra heteroscedasticity, GAMLSS models were proposed including variables to explain the standard error. Moreover, the ARMA and GARMA models were fitted to include the serial correlation structure. The aim of this work is to compare estimates, standard errors, coverage of confidence intervals and mean squared error of predicted value for the various models and choose the most appropriate model, which depends on a complete analysis of residuals, usually omitted in the literature. The GARMA model with Negative Binomial distribution was the best fit since the errors seem to follow the proposed distribution and they have small values of autocorrelation. Besides, this model had low mean squared error and a good coverage of confidence interval. The effect of autocorrelation of data in the estimates was also analyzed in the setting of several models based on simulated data.
|
5 |
Modelos paramétricos para séries temporais de contagem / Parametric models for count time seriesIgor André Milhorança 14 May 2014 (has links)
Diversas situações práticas exigem a análise de series temporais de contagem, que podem apresentar tendência, sazonalidade e efeitos de variáveis explicativas. A motivação do nosso trabalho é a análise de internações diárias por doenças respiratórias para pessoas com mais que 65 anos residentes no município de São Paulo. O efeito de variáveis climáticas e concentrações de poluentes foram incluídos nos modelos e foram usadas as funções seno e cosseno com periodicidade de um ano para explicar o padrão sazonal e obter os efeitos das variáveis climáticas e poluentes controlando essa sazonalidade. Outro aspecto a ser considerado é a inclusão da população nas análises de modo que a interpretação dos efeitos seja para as taxas diárias de internações. Diferentes modelos paramétricos foram propostos para as internações. O mais simples é o modelo de regressão linear para o logaritmo das taxas. Foram ajustados os modelos lineares generalizados (MLG) para as internações com função de ligação logaritmo e com a população como offset, por este modelo permitir o uso das distribuições Poisson e Binomial Negativa, usadas para dados de contagem. Devido à heteroscedasticidade extra, foram propostos modelos GAMLSS incluindo variáveis para explicar o desvio padrão. Foram ajustados modelos ARMA e GARMA, por incluírem uma estrutura de correlação serial. O objetivo desse trabalho é comparar as estimativas, os erros padrões, a cobertura dos intervalos de confiança e o erro quadrático médio para o valor predito segundo os vários modelos e a escolha do modelo mais apropriado, que depende da completa análise de resíduos, geralmente omitida na literatura. O modelo GARMA com distribuição Binomial Negativa apresentou melhor ajuste, pois os erros parecem seguir a distribuição proposta e tem baixa autocorrelação, além de ter tido uma boa cobertura pelo intervalo de confiança e um baixo erro quadrático médio. Também foi analisado o efeito da autocorrelação dos dados nas estimativas nos vários modelos baseado em dados simulados. / Many practical situations require the analysis of time series of counts, which may present trend, seasonality and effects of covariates. The motivation of this work is the analysis of daily hospital admissions for respiratory diseases in people over 65 living in the city of São Paulo. The effect of climatic variables and concentrations of pollutants were included in the models and the sine and cosine functions with annual period were included to explain the seasonal pattern and obtain the effects of pollutants and climatic variables partially controlled by this seasonality. Another aspect to be considered is the inclusion of the population in the analys es in order to interpret the effects based on daily hospitalization rates . Different parametric models have been proposed for hospitalizations. The simplest is the linear regression model for the logarithm of the hospitalization rate. The generalized linear models (GLM) were adjusted for daily admissions with logarithmic link function and the population as offset to consider the Poisson and Negative Binomial distributions for counting data. Due to the extra heteroscedasticity, GAMLSS models were proposed including variables to explain the standard error. Moreover, the ARMA and GARMA models were fitted to include the serial correlation structure. The aim of this work is to compare estimates, standard errors, coverage of confidence intervals and mean squared error of predicted value for the various models and choose the most appropriate model, which depends on a complete analysis of residuals, usually omitted in the literature. The GARMA model with Negative Binomial distribution was the best fit since the errors seem to follow the proposed distribution and they have small values of autocorrelation. Besides, this model had low mean squared error and a good coverage of confidence interval. The effect of autocorrelation of data in the estimates was also analyzed in the setting of several models based on simulated data.
|
6 |
Detecting Major Genes Controlling Robustness of Chicken Body Weight Using Double Generalized Linear ModelsZhang, Liming, Han, Yang January 2010 (has links)
Detecting both the majors genes that control the phenotypic mean and those controlling phenotypic variance has been raised in quantitative trait loci analysis. In order to mapping both kinds of genes, we applied the idea of the classic Haley-Knott regression to double generalized linear models. We performed both kinds of quantitative trait loci detection for a Red Jungle Fowl x White Leghorn F2 intercross using double generalized linear models. It is shown that double generalized linear model is a proper and efficient approach for localizing variance-controlling genes. We compared two models with or without fixed sex effect and prefer including the sex effect in order to reduce the residual variances. We found that different genes might take effect on the body weight at different time as the chicken grows.
|
7 |
Examining the invariance of item and person parameters estimated from multilevel measurement models when distribution of person abilities are non-normalMoyer, Eric 24 September 2013 (has links)
Multilevel measurement models (MMM), an application of hierarchical generalized linear models (HGLM), model the relationship between ability levels estimates and item difficulty parameters, based on examinee responses to items. A benefit of using MMM is the ability to include additional levels in the model to represent a nested data structure, which is common in educational contexts, by using the multilevel framework. Previous research has demonstrated the ability of the one-parameter MMM to accurately recover both item difficulty parameters and examinee ability levels, when using both 2- and 3-level models, under various sample size and test length conditions (Kamata, 1999; Brune, 2011). Parameter invariance of measurement models, that parameter estimates are equivalent regardless of the distribution of the ability levels, is important when the typical assumption of a normal distribution of ability levels in the population may not be correct. An assumption of MMM is that the distribution of examinee abilities, which is represented by the level-2 residuals in the HGLM, is normal. If the distribution of abilities in the population are not normal, as suggested by Micceri (1989), this assumption of MMM is violated, which has been shown to affect the estimation of the level-2 residuals. The current study investigated the parameter invariance of the 2-level 1P-MMM, by examining the accuracy of item difficulty parameter estimates and examinee ability level estimates. Study conditions included the standard normal distribution, as a baseline, and three non-normal distributions having various degrees of skew, in addition to various test lengths and sample sizes, to simulate various testing conditions. The study's results provide evidence for overall parameter invariance of the 2-level 1P-MMM, when accounting for scale indeterminacy from the estimation process, for the study conditions included. Although, the error in the item difficulty parameter and examinee ability level estimates in the study were not of practical importance, there was some evidence that ability distributions may affect the accuracy of parameter estimates for items with difficulties greater than represented in this study. Also, the accuracy of abilities estimates for non-normal distributions seemed less for conditions with greater test lengths and sample sizes, indicating possible increased difficulty in estimating abilities from non-normal distributions. / text
|
8 |
Statistické modely pro kapitálové modely pojišťoven / Statistical models for capital models of insurance companiesŠvagerková, Lýdia January 2011 (has links)
This work deals with the topic of lapse rate modelling in the field of Life Insurance. First, the theoretical apparatus is established: the linear models and their extension, generalized linear models. Furthermore, we describe the process of model selection and evaluation. In the second part of this work we describe the influence of various individual as well as macroeconomical parameters on the lapse rate. We summarize the findings of previous works in this field. The last part introduces models in statistical software R based on generalized linear models and describes the process of their selection and evaluation. Outputs from these models are interpreted and compared to the ratio analysis results.
|
9 |
Gaining Insight With Recursive Partitioning Of Generalized Linear ModelsRusch, Thomas, Zeileis, Achim 06 1900 (has links) (PDF)
Recursive partitioning algorithms separate a feature space into a set of disjoint rectangles.
Then, usually, a constant in every partition is fitted. While this is a simple and
intuitive approach, it may still lack interpretability as to how a specific relationship between dependent and
independent variables may look. Or it may be that a certain model is assumed or of
interest and there is a number of candidate variables that may non-linearily give rise to
different model parameter values.
We present an approach that combines generalized linear models with recursive partitioning
that offers enhanced interpretability of classical trees as well as providing an
explorative way to assess a candidate variable's influence on a parametric model.
This method conducts recursive partitioning of a the generalized linear model by
(1) fitting the model to the data set, (2) testing for parameter instability over a set of
partitioning variables, (3) splitting the data set with respect to the variable associated with
the highest instability. The outcome is a tree where each terminal node is associated with a generalized linear model.
We will show the methods versatility and suitability to gain additional insight
into the relationship of dependent and independent variables by two examples, modelling
voting behaviour and a failure model for debt amortization. / Series: Research Report Series / Department of Statistics and Mathematics
|
10 |
Chasin’ Tail in Southern Alabama: Delineating Programmed and Stimulus-driven Grooming in Odocoileus virginianusHeine, Kyle 11 August 2015 (has links)
This study examined variation in ectoparasite density and grooming behavior of naturally occurring white-tailed deer (Odocoileus virginianus) in southwest Alabama. Stimulus-driven grooming as well as the intraspecific body size and vigilance principles of programmed grooming were tested. During the rut, males had a higher average tick (Ixodidae) density than females and exhibited complete separation of tick parasitism between non-rutting and rutting periods, supporting the vigilance principle. Stimulus-driven grooming was supported, as both fawns and yearlings had significantly higher fly (Hippoboscidae) and combined fly/tick densities than adults, and fawns oral groomed at a significantly higher rate than adults, even in the absence of allogrooming. Programmed and stimulus-driven grooming of deer examined in this study were not mutually exclusive but ectoparasite dependent.
|
Page generated in 0.2464 seconds