Latent class analysis explains dependency structures in multivariate categorical data by assuming the presence of latent classes. We investigate the specification of suitable priors for the Bayesian latent class model to determine the number of classes and perform variable selection. Estimation is possible using standard tools implementing general purpose Markov chain Monte Carlo sampling techniques such as the software JAGS. However, class specific inference requires suitable post-processing in order to eliminate label switching. The proposed Bayesian specification and analysis method is applied to the Hungarian heart disease data set to determine the number of classes and identify relevant variables and results are compared to those obtained with the standard prior for the component specific parameters.

The variation in the cleat permeability of coalbed methane (CBM) reservoirs is attributed primarily to two cardinal processes, with opposing effects. Increase in effective stresses with reduction in pore pressure tends to decrease the cleat permeability, whereas the sorption-induced coal matrix shrinkage actuates reduction in the effective stresses which increases the reservoir permeability. The net effect of the two processes determines the pressure-dependent-permeability and, hence, the overall trend of CBM production with depletion. Several analytical models have been developed and used to predict the dynamic behavior of CBM reservoir permeability during production through pressure depletion, all based on combining the two effects. The purpose of this study was to introduce modifications to two most commonly used permeability models, namely the Palmer and Mansoori, and Shi and Durucan, for permeability variation and evaluate their performance when projecting gas production. The basis for the modification is the linear relationship between the volume of sorbed gas and the associated matrix shrinkage. Hence, the impact of matrix shrinkage is incorporated as a function of the amount of gas produced, or that remaining in coal, at any time during production. Since the exact production from a reservoir is known throughout its life, this significantly simplifies the process of permeability modeling. Furthermore, the modification is also expected to streamline the process of modeling by classifying the shrinkage parameters for coals of different regions, but with similar characteristics. A good analogy is the San Juan basin, where sorption characteristics of coal are so well understood and defined that operators no longer carry out laboratory sorption work. The goal is to achieve the same for incorporation of the matrix shrinkage behavior. Another modification is to incorporate the matrix, or grain, compressibility effect of coal as a correction factor in the Shi and Durucan model so as to assess the permeability variation based on the true shrinkage of coal matrix with reservoir drawdown. Finally, application of the modified models may be carried out for scenarios where the gas content of coal varies with time, either due to injection of a second gas to enhance the recovery of methane, or gas enhancing techniques, such as, bio-stimulation of coal. The original models are currently unable to handle this, particularly when the gas content of the reservoir increases. The research is aimed at simplifying and, in fact, improving the performance of the theoretical models in predicting the variation of coal reservoir permeability.

abstract: The primary objective in time series analysis is forecasting. Raw data often exhibits nonstationary behavior: trends, seasonal cycles, and heteroskedasticity. After data is transformed to a weakly stationary process, autoregressive moving average (ARMA) models may capture the remaining temporal dynamics to improve forecasting. Estimation of ARMA can be performed through regressing current values on previous realizations and proxy innovations. The classic paradigm fails when dynamics are nonlinear; in this case, parametric, regime-switching specifications model changes in level, ARMA dynamics, and volatility, using a finite number of latent states. If the states can be identified using past endogenous or exogenous information, a threshold autoregressive (TAR) or logistic smooth transition autoregressive (LSTAR) model may simplify complex nonlinear associations to conditional weakly stationary processes. For ARMA, TAR, and STAR, order parameters quantify the extent past information is associated with the future. Unfortunately, even if model orders are known a priori, the possibility of over-fitting can lead to sub-optimal forecasting performance. By intentionally overestimating these orders, a linear representation of the full model is exploited and Bayesian regularization can be used to achieve sparsity. Global-local shrinkage priors for AR, MA, and exogenous coefficients are adopted to pull posterior means toward 0 without over-shrinking relevant effects. This dissertation introduces, evaluates, and compares Bayesian techniques that automatically perform model selection and coefficient estimation of ARMA, TAR, and STAR models. Multiple Monte Carlo experiments illustrate the accuracy of these methods in finding the "true" data generating process. Practical applications demonstrate their efficacy in forecasting. / Dissertation/Thesis / Doctoral Dissertation Statistics 2018

La sécheresse, accentuée par le réchauffement climatique actuel et à venir, induit des phénomènes de retrait et de gonflement dans de nombreux sols argileux. Ces phénomènes se traduisent principalement par deux conséquences majeures : la modification des propriétés physiques du sol qui influe directement sur l’agriculture et les déformations induisant souvent des tassements différentiels aux niveaux des structures et des ouvrages. Jusqu’à aujourd’hui, ces phénomènes ont été étudiés principalement à l’échelle mésoscopique (échelle « classique » de laboratoire) et macroscopique (échelle de l’ouvrage). Des recherches ont été menées à l’échelle microscopique mais dans des contextes particuliers (argiles destinées au stockage des déchets radioactifs). Le travail de recherche présenté dans ce mémoire de thèse constitue par conséquent une des premières études sur des argiles naturelles, prélevées in situ sur des sites affectés par le retrait-gonflement : la montmorillonite grecque et l’argile verte de Romainville. Ces deux argiles ont été soumises à des cycles d’humidification-séchage dans le Microscope Electronique à Balayage Environnemental (MEBE) afin de quantifier le retrait-gonflement et d’en étudier la cinétique. Un protocole expérimental a été établi, basé sur l’application de l’humidification et du séchage par paliers avec des temps d’équilibre et une fréquence d’application réguliers. Ces essais ont été menés dans le but d’analyser : l’influence de la composition minéralogique, l’étude de l’effet d’échelle et l’influence du type de pilotage (pression ou température). En parallèle à ces essais, des expérimentations ont été menées à l’échelle mésoscopique afin d’établir d’éventuels liens entre les deux échelles. Ils ont consisté en l’étude de l’évolution de la microstructure, de la microporosité et de l’établissement des courbes de rétention. Les observations effectuées à ces deux échelles ont mené à la mise au point de deux méthodes, volontairement simplifiées pour des facilités de mise en œuvre et d’une investigation rapide (études préliminaires, bureaux d’étude). Ces deux méthodes sont, l’une analytique, l’autre numérique, permettant la transition vers l’échelle macroscopique et pour pouvoir estimer le retrait et le gonflement possibles sur un site donné. L’analyse des résultats des essais effectués à l’échelle microscopique (essais MEBE) conduit aux remarques suivantes : la majeure partie des déformations se produit aux fortes humidités relatives ; les déformations mesurées à l’échelle microscopique sont similaires à celles mesurées à l’échelle mésoscopique ; la cinétique de gonflement et de retrait est marquée par plusieurs phases distinctes selon l’argile considérée ; la composition minéralogique, la taille des agrégats et le type de pilotage présentent une influence au niveau de l’amplitude du retrait-gonflement et de la cinétique. L’analyse des résultats des essais effectués à l’échelle mésoscopique (essais de laboratoire) conduit aux remarques suivantes : la limite de retrait (Wr) correspond à une teneur en eau de 13 % et un indice des vides de 0,52 ; aux fortes et moyennes succions, le sol présente une microstructure compacte avec une macroporosité augmentant avec la succion ; aux faibles succions, la microstructure compacte évolue de contours nets et saillants des agrégats à des contours plus doux et réguliers ; des microorganismes ont été observés, engendrant une porosité à la surface très importante, pouvant entraîner un vieillissement du sol et la création de chemins préférentiels pour la pénétration de l’eau ; avec la diminution de la succion, une diminution de la surface spécifique externe et du volume microporeux est observée ; l’humidification et le séchage des échantillons ne semblent pas induire pas de conséquences irréversibles sur la microstructure. / The drought, emphasised by the current global warming and to come, leads phenomena of shrinkage and swelling of many clayey soils. These phenomena are mainly translated by two major consequences: the modification of the physical properties of the soil which influences directly the agriculture and the deformations leading often differential compaction at the levels of the structures and the buildings. Until today, these phenomena were mainly studied at the mesoscopic level (laboratory classical level) and macroscopic level (buildings scale). Research was led to the microscopic level but in particular contexts (clays intended for the confinement of the radioactive waste). The research work presented in this thesis manuscript constitutes consequently one of the first studies on natural clays, taken in situ from sites affected by the swelling-shrinkage: the Greek montmorillonite and the clay of Romainville. These two clays were subjected to wetting-drying cycles in the Environmental Scanning Electron Microscope (ESEM) in order to quantify the swelling-shrinkage and to study kinetics. An experimental protocol was established, based on the application of the wetting and the drying by stages with regular time of stabilisation and frequency of application. These tests were led in order to analyse: the influence of the mineralogical composition, the study of the effect of scale and the influence of the type of piloting (pressure or temperature). In parallel to these tests, some experiments were led to the mesoscopic level to establish possible links between both levels. They consisted of the study of the evolution of the microstructure, the microporosity and the establishment of the retention curves. The observations made in these two levels led to the development of two methods, voluntarily simplified for ease of implementation and of a fast investigation (preliminary studies, engineering consulting firms). These two methods are, the one analytics, the other numerical, allowing the transition towards the macroscopic level and to be able to estimate the possible shrinkage and the swelling on a given site. The analysis of the results of the tests made in the microscopic level (ESEM tests) leads to the following remarks: the major part of the strains occurs in the high relative humidities; the strains measured in the microscopic level are similar to those measured in the mesoscopic level; the kinetics of swelling and shrinkage are marked by several different phases according to the considered clay; the mineralogical composition, the size of the aggregates and the type of piloting present an influence at the level of the amplitude of the shrinkage - swelling and the kinetics. The analysis of the results of the tests made in the mesoscopic level (laboratory tests) leads to the following remarks: the shrinkage limit (Wr) corresponds to a 13 % moisture content and a void index of 0,52; in the high and medium suctions, the soil presents a compact microstructure with a macroporosity increasing with the suction; in the low suctions, the compact microstructure evolves of sharp outlines of the aggregates in rounder outlines; microorganisms were observed, engendering a very important porosity on the surface, being able to pull an ageing of the ground and the creation of preferential paths for the penetration of the water; with the decrease of the suction, the decrease of the external specific surface and of the microporous volume is observed; the wetting and the drying of samples do not seem to lead no irreversible consequences on the microstructure.

RESUMO Dada sua versatilidade e excelência estética, a resina composta é o material restaurador direto mais utilizado nas intervenções da clínica odontológica. Uso tão intenso levou pesquisadores e fabricantes a trabalharem na evolução deste material, conferindo-lhe aumento em sua resistência à compressão e abrasão, bem como melhora em seu comportamento estético. Infelizmente, um problema intrínseco da resina, a contração de polimerização é sua principal desvantagem, e caracteriza-se pela alteração volumétrica que sofre quando as moléculas de seus monômeros se unem para a formação das cadeias poliméricas. Uma das principais técnicas para combater os efeitos nocivos desta alteração volumétrica é a inserção incremental do material na cavidade a ser restaurada. Nesta pesquisa, a contração de polimerização de uma resina composta de uso universal (Charisma Heraeus Kullser) será avaliada variando-se a técnica de inserção do material na cavidade em incremento único ou três incrementos oblíquos, através do método da picnometria a gás. O picnômetro Accupyc 1330 (Micromeritics) permite medições rápidas e fiéis do volume da resina composta antes e após a fotoativação. Os resultados mostram que ocorreram diferenças estatísticas entre as duas técnicas, sendo a técnica de único incremento a que apresentou menor contração de polimerização. / Because of its versatility and aestetic excellency, composite resin is one of the most utilized direct restorative materials. Because of the resin intense use, the number of researches to improve its performance are numerous, and can be verified by the resin aestethetic behavior and its compressive and abrasion strenght increase. In spite of the mentioned developments, the polymerization shrinkage inherent of the material had been a major deficiency, and can be understood by the alteration in volume of the material after the polymerization process. In the present study, the polymerization shrinkage of the restorative material Charisma (Heraes Kulzer), was evaluated using bulk and incremental placement techniques and measured by gas pycnometer. Accupyc 1330 (Micromeritics, USA) pycnometer measures the helium displacement with great precision, allowing fast and accurate measurements of the volume of the composite resin before and after polymerization, without interference of the moisture or temperature. The results showed statistically significant diferences between the two placement techniques, with the bulk technique showing lower polymerization shrinkage than the incremental technique.

Recent advances in technology for brain imaging and high-throughput genotyping have motivated studies examining the influence of genetic variation on brain structure. In this setting, high-dimensional regression for multi-SNP association analysis is challenging as the brain imaging phenotypes are multivariate and there is a desire to incorporate a biological group structure among SNPs based on their belonging genes. Wang et al. (Bioinformatics, 2012) have recently developed an approach for simultaneous estimation and SNP selection based on penalized regression with regularization based on a novel group l_{2,1}-norm penalty, which encourages sparsity at the gene level. A problem with the proposed approach is that it only provides a point estimate. We solve this problem by developing a corresponding Bayesian formulation based on a three-level hierarchical model that allows for full posterior inference using Gibbs sampling. For the selection of tuning parameters, we consider techniques based on: (i) a fully Bayes approach with hyperpriors, (ii) empirical Bayes with implementation based on a Monte Carlo EM algorithm, and (iii) cross-validation (CV). When the number of SNPs is greater than the number of observations we find that both the fully Bayes and empirical Bayes approaches overestimate the tuning parameters, leading to overshrinkage of regression coefficients. To understand this problem we derive an approximation to the marginal likelihood and investigate its shape under different settings. Our investigation sheds some light on the problem and suggests the use of cross-validation or its approximation with WAIC (Watanabe, 2010) when the number of SNPs is relatively large. Properties of our Gibbs-WAIC approach are investigated using a simulation study and we apply the methodology to a large dataset collected as part of the Alzheimer's Disease Neuroimaging Initiative. / Graduate

The use of supplementary cementitious materials (SCMs) in commercial construction have been increasing over the last several decades as climate change and sustainability has been gaining global attention. Incorporation of SCMs into concrete mixtures provides several environmental benefits. Since most SCMs are waste by-products of other industries, their use in concrete reduces waste disposal. Additionally, cements substitution with SCMs reduces the carbon footprint of the construction industry. Cement production generates large amounts of CO2 emissions; the use of SCMs reduces the amount of cement in a concrete mixture thereby reducing emissions from its production. In addition to SCMs, modern concretes typically incorporate a combination of chemical admixtures. Adverse interaction of admixtures with cement, with or without the SCMs, or with each other is one of the most common reasons for early-age concrete issues. Since the possible combinations of admixtures are numerous and there is a variety of cements on the market, testing all possible chemical/mineral/cement admixture combinations is impractical. The aim of this research was to cover a broad base of admixture-related issues, each addressing a specific need of the construction industry. There is currently no explanation for why calcium chloride-based accelerator is not always effective when used with high tricalcium aluminate (C3A) cements. It was determined that increasing C3A or gypsum content alone did not appear to significantly affect acceleration; however, the presence of alkalis reduced the effectiveness of CaCl2 accelerator. When CaCl2-based accelerators are used in concrete, they are typically used in combination with other chemical admixtures, such as water-reducing and retarding admixtures (WRRA) to allow for the use of a low water-cementitious material ratio. In order to avoid premature hardening, CaCl2 accelerator is most often added onsite, rather than at the concrete batching plant. Onsite addition can lead to accidental overdose of accelerator. It was found that increasing dosages of calcium chloride-containing accelerating admixtures in the presence of WRRA has a non-linear effect on the pore size distribution and consequently a non-linear increase on the autogenous shrinkage, which can contribute to early-age concrete cracking. Water-reducing admixtures and superplasticizers are added to concrete to improve workability, which decreases not only with a decrease in water-cementitious material ratio, but also with addition of some SCMs. Silica fume and metakaolin are known to decrease workability; fly ash and slag addition improve it. The effect of SCM combinations on workability is typically assumed to be additive. However, this investigation revealed that combining SCMs does not have an additive effect on workability, measured in terms of apparent yield stress and plastic viscosity; consequently, these parameters cannot be estimated from their respective values. Cement replacement with SCMs affects not only workability, but also heat of hydration, and is commonly used to reduce concrete temperature rise in concrete. Prediction and control of concrete temperature rise due to cement hydration is of great significance for mass concrete structures since large temperature gradients between the surface and the core of the structure can lead to cracking thus reducing durability of the structure. A number of equations have been proposed to predict the heat of hydration of cement and cement/SCM blends. However, these equations do not include metakaolin, which is a relatively new mineral admixture. Based on statistical experimental design, an equation was developed to predict the reduction of total hydration heat at 24, 48 and 72 hours with addition of SCMs compared to a plain ordinary portland cement (OPC)-water mixture. The developed equation allows the evaluation of the contribution of Class F fly ash (FA), blast furnace slag (BFS), silica fume (SF) and metakaolin (MK) as well as their combinations. Since metakaolin has been on the market for only about 10 years, the current knowledge on its effect on hydration products and paste microstructure remains incomplete. The effect of MK on the nature of hydration products was evaluated through x-ray diffraction. Its effect on the microstructure was assessed by measuring porosity with nitrogen adsorption and determining nanoindentation modulus as well as the volume fraction of calcium silicate hydrates (C-S-H) with variable packing densities. No significant effect was observed on the nature of hydration products with MK or BFS addition. However, nitrogen-accessible porosity increased with MK and BFS addition, the increase being larger with BFS. The average indentation modulus for the hydration products decreased with addition of MK and BFS, which corresponded to increasing nitrogen accessible pores. The results of this study indicate that phase quantification by quantitative x-ray diffraction (QXRD) of the hydrated paste may not be sufficient to assess the impact of metakaolin or BFS addition on the hydrating cementitious systems, and a multi-technique approach that provides information not only on the amount of hydration products, but also their morphology is preferable.

Ce travail a pour objectif d’étudier la fissuration des sols due au séchage par des approches numérique et analytique. L’initiation et la propagation des fissures sont investiguées en utilisant un code de calcul aux éléments finis avec la présence des joints cohésifs. Les couplages entre le problème hydraulique et le comportement mécanique en présence des discontinuités sont considérés. La loi de la fissure cohésive est appliquée pour modéliser l’initiation et la propagation des fissures.Tout d’abord, les résultats d’un essai de séchage au laboratoire réalisé sur un sol argileux à l’état liquide sont utilisés afin d’évaluer la méthode numérique proposée. Les résultats numériques montrent que le modèle est capable de reproduire les tendances principales du processus de séchage. Elle souligne aussi l’importance des conditions aux limites dans l’initiation des fissures. Ensuite, une approche énergétique est proposée pour étudier l’initiation d’une fissure. Les énergies élastiques avant et après l’initiation de la fissure sont estimées par les deux approches analytique et numérique. L’énergie dissipée lors de l’initiation de la fissure est comparée avec le taux d’énergie pour créer une fissure. Les analyses montrent que le critère d’énergie peut est atteint avant le critère de contrainte. La dissipation de l’énergie cumulée correspond à la propagation instable lors de l’initiation de la fissure. De plus, le développement et la géométrie des fissures sont étudiés essentiellement par les simulations numériques avec plusieurs joints cohésifs. Les résultats numériques montrent que la fissuration se produit souvent progressivement pour former différentes familles de fissures par un processus dichotomique (une fissure apparait au milieu de deux fissures existantes). La propagation d’une fissure est brutale dans la phase d’initiation pour atteindre une profondeur appelée ‘ultime’. Les fissures dans chaque famille peuvent apparaitre simultanément à un même niveau de succion et présentent une même profondeur ultime. En se basant sur les résultats numériques et quelques analyses analytiques supplémentaires, des relations empiriques sont proposées afin de prédire l’espacement et la profondeur ultime des fissures. Finalement, quelques calculs préliminaires sont réalisés afin d’évaluer le potentiel de la méthode numérique proposée pour prédire la fissuration liée au séchage des ouvrages en terre / This work focuses on the analysis of desiccation cracking by numerical and analytical approaches. The initiation and the propagation of cracks are investigated by using a finite element code including cohesive joints elements. Coupling between the hydraulic and the mechanical processes in the presence of discontinuities is considered. The cohesive crack’s law is applied to simulate the initiation and the propagation of cracks.Results of a laboratory experiments performed on slurry clay soil are first used to evaluate the proposed numerical modelling method. The results show that the method is able to reproduce the main trends of desiccation process. The importance of boundary conditions are also discussed. Second, an energy approach is proposed to study the initiation of cracks. The energies before and after crack initiation are estimated by both numerical and analytical solutions. The energy released by cracking is then compared to the crack energy to discuss crack initiation conditions. The analysis shows that the energy criterion is reached before the stress criterion, and this can explain unstable crack propagation at the beginning. Third, the development and the geometry of desiccation cracks are studied by numerical simulation with several cohesive joints. The numerical results show that cracking occurs sequentially to form different cracks families with a dichotomy process (the subsequent cracks appear at the middle of two existing neighboring ones). The cracks in each family appear simultaneously and reach an identical ultimate depth. From the numerical results and additional analytical analysis, empirical correlations are proposed to predict the spacing and crack depth. Finally, some preliminary studies are performed showing that the proposed numerical method can be used to predict the desiccation crack phenomena observed on geotechnical earth structures

Volume changes are one of the priority characteristics of concrete that plays an important role place mainly in the durability of structures as such. These low volume changes are required especially for structures as: industrial concrete floors, massive structures and waterproof structures. One of the possibilities in reducing these changes is the use of mineral admixtures with suitable grain curve of aggregate. Another possibility is the use of shrinkage reducing admixtures. There can occur problem with efficiency (long term age) with other components of the mixture. Therefore, the effort of this work is to use the avaible processes for maximum reduction of shrinkage and hydration temperatures.

Indiana University-Purdue University Indianapolis (IUPUI) / Process induced dimensional changes in composite parts has been the topic of interest for many researchers. The residual stresses that are induced in composite laminates during curing process while the laminate is in contact with the process tool often lead to dimensional variations such as spring-in of angles and warpage of flat panels. The traditional trial-and-error approach can work for simple geometries, but composite parts with complex shapes require more sophisticated models. When composite laminates are subjected to thermal stresses, such as the heating and cooling processes during curing, they can become distorted as the in-plane and the throughthickness coeffcients of thermal expansion are di erent, as well as chemical shrinkage of the resin, usually cause spring-in. Deformed components can cause problems during assembly, which significantly increases production costs and affects performance. This thesis focuses on predicting these shape deformations using software simulation of composite manufacturing and curing. Various factors such as resin shrinkage, degrees of cure, difference between through thickness coeffcient of thermal expansion of the composite laminate are taken into the consideration. A cure kinetic model is presented which illustrates the matrix behavior during cure. The results obtained using the software then were compared with the experimental values of spring-in from the available literature. The accuracy of ACCS package was validated in this study. Analyzing the effects of various parameters of it was estimated that 3D part simulation is an effective and cost and time saving method to predict nal shape of the composite part.