The development of academic English proficiency and the time it takes to reclassify to fluent English proficient status are key issues in monitoring achievement of English learners. Yet, little is known about academic English language development at the domain-level (listening, speaking, reading, and writing), or how English language development is associated with time-to-reclassification as an English proficient student. Although the substantive findings surrounding English proficiency and reclassification are of great import, the main focus of this dissertation was methodological: the exploration and testing of joint modeling methods for studying both issues. The first joint model studied was a multilevel, multivariate random effects model that estimated the student-specific and school-specific association between different domains of English language proficiency. The second model was a multilevel shared random effects model that estimated English proficiency development and time-to-reclassification simultaneously and treated the student-specific random effects as latent covariates in the time-to-reclassification model. These joint modeling approaches were illustrated using annual English language proficiency test scores and time-to-reclassification data from a large Arizona school district. Results from the multivariate random effects model revealed correlations greater than .5 among the reading, writing and oral English proficiency random intercepts. The analysis of English proficiency development illustrated that some students had attained proficiency in particular domains at different times, and that some students had not attained proficiency in a particular domain even when their total English proficiency score met the state benchmark for proficiency. These more specific domain score analyses highlight important differences in language development that may have implications for instruction and policy. The shared random effects model resulted in predictions of time-to-reclassification that were 97% accurate compared to 80\% accuracy from a conventional discrete-time hazard model. The time-to-reclassification analysis suggested that use of information about English language development is critical for making accurate predictions of the time a student will reclassify in this Arizona school district.

No description available.

No description available.

La mucoviscidose est à ce jour une maladie malheureusement incurable. Malgré les nombreux progrès réalisés dans la recherche à ce sujet, il reste indispensable d’avoir davantage une meilleure connaissance de la maladie afin de proposer des traitements encore plus adaptés aux patients. La majorité des traitements actuels visent principalement à réduire les symptômes de la maladie sans toutefois la guérir. A ce jour, la transplantation pulmonaire reste le moyen le plus adéquat pour améliorer la qualité de vie et prolonger la vie des patients dont l’état respiratoire s’est considérablement dégradé. Cependant les critères d’identification des patients éligibles à la transplantation pulmonaire sont assez variés. Il est donc nécessaire de fournir aux cliniciens des outils d’aide à la décision pour mieux identifier les patients nécessitant une transplantation pulmonaire. Pour ce faire, il est indispensable de connaitre d’une part, les facteurs pronostics de transplantation pulmonaire et d’autre part, de savoir convenablement pronostiquer la survenue de cet événement chez les sujets atteints de mucoviscidose. L’objectif de ce travail de thèse est de développer des outils pronostiques utiles à l’évaluation des choix thérapeutiques liés à la transplantation pulmonaire. Dans la première partie de ce travail, nous avons réévalué les facteurs pronostiques de la transplantation pulmonaire ou du décès chez les adultes atteints de mucoviscidose. Suite aux progrès thérapeutiques qui ont conduit à l’amélioration du pronostic au cours des dernières années, ce travail a permis d’identifier des facteurs pronostiques en adéquation avec l’état actuel de la recherche. Un deuxième travail a consisté à développer un modèle conjoint à classes latentes fournissant des prédictions dynamiques pour la transplantation pulmonaire ou le décès. Ce modèle a permis d’identifier trois profils d’évolution de la maladie et également d’actualiser le risque de survenue de la transplantation pulmonaire ou du décès à partir des données longitudinales du marqueur VEMS. Ces modèles pronostiques ont été développés à partir des données du registre français de la mucoviscidose et ont fourni de bonnes capacités prédictives en termes de discrimination et de calibration. / Cystic Fibrosis is unfortunately an incurable inherited disorder. Despite real progress in research, it is essential to always have a better understanding of the disease in order to provide suitable treatments to patients. Current treatments mostly aim to reduce the disease symptoms without curing it. Lung transplantation is proposed to cystic fibrosis patients with terminal respiratory failure with the aim of improving life expectancy and quality of life. However, criteria for referring patients for lung transplantation still vary widely among transplant centers. It is necessary to guide clinicians in identifying in a good way patients requiring an evaluation for lung transplantation. It is thus important to clearly identify prognostic factors related to lung transplantation and to predict in a good way the occurrence of this event in patients with cystic fibrosis. The aim of this work was to develop prognostic tools to assist clinicians in the evaluation of different therapeutic options related to lung transplantation. First, we reevaluated prognostic factors of lung transplantation or death in adult with cystic fibrosis. indeed, therapeutic progress in patients with cystic fibrosis has resulted in improved prognosis over the past decades. We identified prognostic factors related to the current state of research in the cystic fibrosis field. We further developed a joint model with latent classes which provided dynamic predictions for lung transplantation or death. This model identified three profile of the evolution of the disease and was able to update the risk of lung transplantation or death taking into account the evolution of the longitudinal marker FEV1 which describes the lung function. These prognostic models were developed using the French cystic fibrosis registry and provided good predictive accuracies in terms of discrimination and calibration.

Joint models for the association of a longitudinal binary and a longitudinal continuous process are proposed for situations in which their association is of direct interest. The models are parameterized such that the dependence between the two processes is characterized by unconstrained regression coefficients. Bayesian variable selection techniques are used to parsimoniously model these coefficients. A Markov chain Monte Carlo (MCMC) sampling algorithm is developed for sampling from the posterior distribution, using data augmentation steps to handle missing data. Several technical issues are addressed to implement the MCMC algorithm efficiently. The models are motivated by, and are used for, the analysis of a smoking cessation clinical trial in which an important question of interest was the effect of the (exercise) treatment on the relationship between smoking cessation and weight gain.

Data from insect studies may present different features. Univariate responses may be analyzed using generalized linear models (continuous and discrete data), survival models (time until event data), mixed effects models (longitudinal data), among other methods. These models may be used to analyse data from experiments which assess complex ecological processes, such as competition and predation. In that sense, computational tools are useful for researchers in several fields, e.g., insect biology and physiology, applied ecology and biological control. Using different datasets from entomology as motivation, as well as other types of datasets for illustration purposes, this work intended to develop new modelling frameworks and goodness-of-fit assessment tools. We propose accelerated failure rate mixed models with simultaneous location and scale modelling with regressors to analyse time-until-attack data from a choice test experiment. We use the exponential, Weibull and exponentiated-Weibull models, and assess goodness-of-fit using half-normal plots with simulation envelopes. These plots are the subject of an entire Chapter on an R package, called hnp, developed to implement them. We use datasets from different types of experiments to illustrate the use of these plots and the package. A bivariate extension to the N-mixture modelling framework is proposed to analyse longitudinal count data for two species from the same food web that may interact directly or indirectly, and example datasets from ecological studies are used. An advantage of this modelling framework is the computation of an asymmetric correlation coefficient, which may be used by ecologists to study the degree of association between species. The jointNmix R package was also developed to implement the estimation process for these models. Finally, we propose a goodness-of-fit assessment tool for bivariate models, analogous to the half-normal plot with a simulation envelope, and illustrate the approach with simulated data and insect competition data. This tool is also implemented in an R package, called bivrp. All software developed in this thesis is made available freely on the Comprehensive R Archive Network. / Dados provenientes de estudos com insetos podem apresentar características diferentes. Respostas univariadas podem ser analisadas utilizando-se modelos lineares generalizados (dados contínuos e discretos), modelos de análise de sobrevivência (dados de tempo até ocorrência de um evento), modelos de efeitos mistos (dados longitudinais), dentre outros métodos. Esses modelos podem ser usados para analisar dados provenientes de experimentos que avaliam processos ecológicos complexos, como competição e predação. Nesse sentido, ferramentas computacionais são úteis para pesquisadores em diversos campos, por exemplo, biologia e fisiologia de insetos, ecologia aplicada e controle biológico. Utilizando diferentes conjuntos de dados entomológicos como motivação, assim como outros tipos de dados para ilustrar os métodos, este trabalho teve como objetivos desenvolver novos modelos e ferramentas para avaliar a qualidade do ajuste. Foram propostos modelos de tempo de vida acelerado mistos, com modelagem simultânea dos parâmetros de locação e de escala com regressores, para analisar dados de tempo até ataque de um experimento que avaliou escolha de predadores. Foram utilizados modelos exponencial, Weibull e Weibull-exponenciado, e a qualidade do ajuste foi avaliada utilizando gráficos meio-normais com envelope de simulação. Esses gráficos são o assunto de um Capítulo inteiro sobre um pacote para o software R, chamado hnp, desenvolvido para implementá-los. Foram utilizados conjuntos de dados de diferentes tipos de experimentos para ilustrar o uso desses gráficos e do pacote. Uma extensão bivariada para os modelos chamados \"N-mixture\" foi proposta para analisar dados longitudinais de contagem para duas espécies pertencentes à mesma teia trófica, que podem interagir direta e indiretamente, e conjuntos de dados provenientes de estudos ecológicos são usados para ilustrar a abordagem. Uma vantagem dessa estratégica de modelagem é a obtenção de um coeficiente de correlação assimétrico, que pode ser utilizado por ecologistas para inferir acerca do grau de associação entre espécies. O pacote jointNmix foi desenvolvido para implemetar o processo de estimação para esses modelos. Finalmente, foi proposta uma ferramenta de avaliação de qualidade do ajuste para modelos bivariados, análoga ao gráfico meio-normal com envelope de simulação, e a metodologia _e ilustrada com dados simulados e dados de competição de insetos. Essa ferramenta está também implementada em um pacote para o R, chamado bivrp. Todo o software desenvolvido nesta tese está disponível, gratuitamente, na Comprehensive R Archive Network (CRAN).

In HIV/AIDS studies, viral load (the number of copies of HIV-1 RNA) and CD4 cell counts are important biomarkers of the severity of viral infection, disease progression, and treatment evaluation. Recently, joint models, which have the capability on the bias reduction and estimates' efficiency improvement, have been developed to assess the longitudinal process, survival process, and the relationship between them simultaneously. However, the majority of the joint models are based on mean regression, which concentrates only on the mean effect of outcome variable conditional on certain covariates. In fact, in HIV/AIDS research, the mean effect may not always be of interest. Additionally, if obvious outliers or heavy tails exist, mean regression model may lead to non-robust results. Moreover, due to some data features, like left-censoring caused by the limit of detection (LOD), covariates with measurement errors and skewness, analysis of such complicated longitudinal and survival data still poses many challenges. Ignoring these data features may result in biased inference. Compared to the mean regression model, quantile regression (QR) model belongs to a robust model family, which can give a full scan of covariate effect at different quantiles of the response, and may be more robust to extreme values. Also, QR is more flexible, since the distribution of the outcome does not need to be strictly specified as certain parametric assumptions. These advantages make QR be receiving increasing attention in diverse areas. To the best of our knowledge, few study focuses on the QR-based joint models and applies to longitudinal-survival data with multiple features. Thus, in this dissertation research, we firstly developed three QR-based joint models via Bayesian inferential approach, including: (i) QR-based nonlinear mixed-effects joint models for longitudinal-survival data with multiple features; (ii) QR-based partially linear mixed-effects joint models for longitudinal data with multiple features; (iii) QR-based partially linear mixed-effects joint models for longitudinal-survival data with multiple features. The proposed joint models are applied to analyze the Multicenter AIDS Cohort Study (MACS) data. Simulation studies are also implemented to assess the performance of the proposed methods under different scenarios. Although this is a biostatistical methodology study, some interesting clinical findings are also discovered.

Data from insect studies may present different features. Univariate responses may be analyzed using generalized linear models (continuous and discrete data), survival models (time until event data), mixed effects models (longitudinal data), among other methods. These models may be used to analyse data from experiments which assess complex ecological processes, such as competition and predation. In that sense, computational tools are useful for researchers in several fields, e.g., insect biology and physiology, applied ecology and biological control. Using different datasets from entomology as motivation, as well as other types of datasets for illustration purposes, this work intended to develop new modelling frameworks and goodness-of-fit assessment tools. We propose accelerated failure rate mixed models with simultaneous location and scale modelling with regressors to analyse time-until-attack data from a choice test experiment. We use the exponential, Weibull and exponentiated-Weibull models, and assess goodness-of-fit using half-normal plots with simulation envelopes. These plots are the subject of an entire Chapter on an R package, called hnp, developed to implement them. We use datasets from different types of experiments to illustrate the use of these plots and the package. A bivariate extension to the N-mixture modelling framework is proposed to analyse longitudinal count data for two species from the same food web that may interact directly or indirectly, and example datasets from ecological studies are used. An advantage of this modelling framework is the computation of an asymmetric correlation coefficient, which may be used by ecologists to study the degree of association between species. The jointNmix R package was also developed to implement the estimation process for these models. Finally, we propose a goodness-of-fit assessment tool for bivariate models, analogous to the half-normal plot with a simulation envelope, and illustrate the approach with simulated data and insect competition data. This tool is also implemented in an R package, called bivrp. All software developed in this thesis is made available freely on the Comprehensive R Archive Network. / Dados provenientes de estudos com insetos podem apresentar características diferentes. Respostas univariadas podem ser analisadas utilizando-se modelos lineares generalizados (dados contínuos e discretos), modelos de análise de sobrevivência (dados de tempo até ocorrência de um evento), modelos de efeitos mistos (dados longitudinais), dentre outros métodos. Esses modelos podem ser usados para analisar dados provenientes de experimentos que avaliam processos ecológicos complexos, como competição e predação. Nesse sentido, ferramentas computacionais são úteis para pesquisadores em diversos campos, por exemplo, biologia e fisiologia de insetos, ecologia aplicada e controle biológico. Utilizando diferentes conjuntos de dados entomológicos como motivação, assim como outros tipos de dados para ilustrar os métodos, este trabalho teve como objetivos desenvolver novos modelos e ferramentas para avaliar a qualidade do ajuste. Foram propostos modelos de tempo de vida acelerado mistos, com modelagem simultânea dos parâmetros de locação e de escala com regressores, para analisar dados de tempo até ataque de um experimento que avaliou escolha de predadores. Foram utilizados modelos exponencial, Weibull e Weibull-exponenciado, e a qualidade do ajuste foi avaliada utilizando gráficos meio-normais com envelope de simulação. Esses gráficos são o assunto de um Capítulo inteiro sobre um pacote para o software R, chamado hnp, desenvolvido para implementá-los. Foram utilizados conjuntos de dados de diferentes tipos de experimentos para ilustrar o uso desses gráficos e do pacote. Uma extensão bivariada para os modelos chamados \"N-mixture\" foi proposta para analisar dados longitudinais de contagem para duas espécies pertencentes à mesma teia trófica, que podem interagir direta e indiretamente, e conjuntos de dados provenientes de estudos ecológicos são usados para ilustrar a abordagem. Uma vantagem dessa estratégica de modelagem é a obtenção de um coeficiente de correlação assimétrico, que pode ser utilizado por ecologistas para inferir acerca do grau de associação entre espécies. O pacote jointNmix foi desenvolvido para implemetar o processo de estimação para esses modelos. Finalmente, foi proposta uma ferramenta de avaliação de qualidade do ajuste para modelos bivariados, análoga ao gráfico meio-normal com envelope de simulação, e a metodologia _e ilustrada com dados simulados e dados de competição de insetos. Essa ferramenta está também implementada em um pacote para o R, chamado bivrp. Todo o software desenvolvido nesta tese está disponível, gratuitamente, na Comprehensive R Archive Network (CRAN).

Cette thèse est consacrée au développement de nouvelles méthodologies pour l'analyse des modèles non-linéaires à effets mixtes, à leur implémentation dans un logiciel accessible et leur application à des problèmes réels. Nous considérons particulièrement des extensions des modèles non-linéaires à effets mixtes aux modèles de mélange et aux modèles conjoints. Dans la première partie, nous proposons, dans le but d'avoir une meilleure maîtrise de l'hétérogénéité liée aux données sur des patients issus de plusieurs clusters, des extensions des MNLEM aux modèles de mélange. Nous proposons ensuite de combiner l'algorithme EM, utilisé traditionnellement pour les modèles de mélanges lorsque les variables étudiées sont observées, et l'algorithme SAEM, utilisé pour l'estimation de paramètres par maximum de vraisemblance lorsque ces variables ne sont pas observées. La procédure résultante, dénommée MSAEM, permet ainsi d'éviter l'introduction d'une étape de simulation des covariables catégorielles latentes dans l'algorithme d'estimation. Cet algorithme est extrêmement rapide, très peu sensible à l'initialisation des paramètres, converge vers un maximum (local) de la vraisemblance et est implémenté dans le logiciel Monolix.La seconde partie de cette Thèse traite de la modélisation conjointe de l'évolution d'un marqueur biologique au cours du temps et les délais entre les apparitions successives censurées d'un évènement d'intérêt. Nous considérons entre autres, les censures à droite, les multiples censures par intervalle d'évènements répétés. Les paramètres du modèle conjoint résultant sont estimés en maximisant la vraisemblance jointe exacte par un algorithme de type MCMC-SAEM. Cette méthodologie est désormais disponible sous Monolix / The main goal of this thesis is to develop new methodologies for the analysis of non linear mixed-effects models, along with their implementation in accessible software and their application to real problems. We consider particularly extensions of non-linear mixed effects model to mixture models and joint models. The study of these two extensions is the essence of the work done in this document, which can be divided into two major parts. In the first part, we propose, in order to have a better control of heterogeneity linked to data of patient issued from several clusters, extensions of NLMEM to mixture models. We suggest in this Thesis to combine the EM algorithm, traditionally used for mixtures models when the variables studied are observed, and the SAEM algorithm, used to estimate the maximum likelihood parameters when these variables are not observed. The resulting procedure, referred MSAEM, allows avoiding the introduction of a simulation step of the latent categorical covariates in the estimation algorithm. This algorithm appears to be extremely fast, very little sensitive to parameters initialization and converges to a (local) maximum of the likelihood. This methodology is now available under the Monolix software. The second part of this thesis deals with the joint modeling of the evolution of a biomarker over time and the time between successive appearances of a possibly censored event of interest. We consider among other, the right censoring and interval censorship of multiple events. The parameters of the resulting joint model are estimated by maximizing the exact joint likelihood by using a MCMC-SAEM algorithm. The proposed methodology is now available under Monolix.

Ce travail a pour but de développer des outils statistiques pour l'étude du déclin cognitif général ou précédant le diagnostic de démence, à partir de données de cohorte en tenant compte du risque compétitif de décès et de la censure par intervalle. Le temps de démence est censuré par intervalle dans les études de cohortes car le diagnostic de démence ne peut être établi qu'à l'occasion des visites qui peuvent être espacées de plusieurs années. Ceci induit une sous-estimation du risque de démence à cause du risque compétitif de décès : les sujets déments sont à fort risque de mourir, et peuvent donc décéder avant la visite de diagnostic. Dans la première partie, nous proposons un modèle conjoint à classes latentes pour données longitudinales corrélées à un événement censuré par intervalle, en compétition avec le décès. Appliqué à la cohorte Paquid, ce modèle permet d'identifier des profils de déclin cognitif associés à des risques différents de démence et de décès. En utilisant cette méthodologie, nous comparons ensuite des modèles pronostiques dynamiques pour la démence, traitant la censure par intervalle, basés sur des mesures répétées de marqueurs cognitifs. Dans la seconde partie, nous conduisons une étude comparative afin de clarifier l'interprétation des estimateurs du maximum de vraisemblance des modèles mixtes et conjoints et estimateurs par équations d'estimation généralisées (GEE), couramment utilisés dans le contexte de données longitudinales incomplètes et tronquées par le décès. Les estimateurs de maximum de vraisemblance ciblent le changement individuel chez les individus vivants. Les estimateurs GEE avec matrice de corrélation de travail indépendante, pondérés par l'inverse de la probabilité d'être observé sachant que le sujet est vivant, ciblent la trajectoire moyennée sur la population des survivants à chaque âge. Ces résultats justifient l'utilisation des modèles conjoints dans l'étude de la démence, qui sont des outils prometteurs pour mieux comprendre l'histoire naturelle de la maladie / The purpose of this work is to develop statistical tools to study the general or the prediagnosis cognitive decline, while accounting for the selection by death and interval censoring. In cohort studies, the time-to-dementia-onset is interval-censored as the dementia status is assessed intermittently. This issue can lead to an under-estimation of the risk of dementia, due to the competing risk of death: subjects with dementia are at high risk to die and can thus die prior to the diagnosis visit. First, we propose a joint latent class illness-death model for longitudinal data correlated to an interval-censored time-to-event, competing with the time-to-death. This model is applied on the Paquid cohort to identify profiles of pre-dementia cognitive declines associated with different risks of dementia and death. Using this methodology, we compare dynamic prognostic models for dementia based on repeated measures of cognitive markers, accounting for interval censoring. Secondly, we conduct a simulation study to clarify the interpretation of maximum likelihood estimators of joint and mixed models as well as GEE estimators, frequently used to handle incomplete longitudinal data truncated by death. Maximum likelihood estimators target the individual change among the subjects currently alive. GEE estimators with independent working correlation matrix, weighted by the inverse probability to be observed given that the subject is alive, target the population-averaged change among the dynamic population of survivors. These results justify the use of joint models in dementia studies, which are promising statistical tools to better understand the natural history of dementia