The problem in this thesis was originally motivated by problems presented with documents of Early England Data Set (DEEDS). The central problem with these medieval documents is the lack of methods to assign accurate dates to those documents which bear no date. With the problems of the DEEDS documents in mind, we present two methods to impute missing features of texts. In the first method, we suggest a new class of metrics for measuring distances between texts. We then show how to combine the distances between the texts using statistical smoothing. This method can be adapted to settings where the features of the texts are ordered or unordered categoricals (as in the case of, for example, authorship assignment problems). In the second method, we estimate the probability of occurrences of words in texts using nonparametric regression techniques of local polynomial fitting with kernel weight to generalized linear models. We combine the estimated probability of occurrences of words of a text to estimate the probability of occurrence of a text as a function of its feature -- the feature in this case being the date in which the text is written. The application and results of our methods to the DEEDS documents are presented.

In this thesis, we address issues of model estimation for longitudinal categorical data and of model selection for these data with missing covariates. Longitudinal survey data capture the responses of each subject repeatedly through time, allowing for the separation of variation in the measured variable of interest across time for one subject from the variation in that variable among all subjects. Questions concerning persistence, patterns of structure, interaction of events and stability of multivariate relationships can be answered through longitudinal data analysis. Longitudinal data require special statistical methods because they must take into account the correlation between observations recorded on one subject. A further complication in analysing longitudinal data is accounting for the non- response or drop-out process. Potentially, the missing values are correlated with variables under study and hence cannot be totally excluded. Firstly, we investigate a Bayesian hierarchical model for the analysis of categorical longitudinal data from the Longitudinal Survey of Immigrants to Australia. Data for each subject is observed on three separate occasions, or waves, of the survey. One of the features of the data set is that observations for some variables are missing for at least one wave. A model for the employment status of immigrants is developed by introducing, at the first stage of a hierarchical model, a multinomial model for the response and then subsequent terms are introduced to explain wave and subject effects. To estimate the model, we use the Gibbs sampler, which allows missing data for both the response and explanatory variables to be imputed at each iteration of the algorithm, given some appropriate prior distributions. After accounting for significant covariate effects in the model, results show that the relative probability of remaining unemployed diminished with time following arrival in Australia. Secondly, we examine the Bayesian model selection techniques of the Bayes factor and Deviance Information Criterion for our regression models with miss- ing covariates. Computing Bayes factors involve computing the often complex marginal likelihood p(y|model) and various authors have presented methods to estimate this quantity. Here, we take the approach of path sampling via power posteriors (Friel and Pettitt, 2006). The appeal of this method is that for hierarchical regression models with missing covariates, a common occurrence in longitudinal data analysis, it is straightforward to calculate and interpret since integration over all parameters, including the imputed missing covariates and the random effects, is carried out automatically with minimal added complexi- ties of modelling or computation. We apply this technique to compare models for the employment status of immigrants to Australia. Finally, we also develop a model choice criterion based on the Deviance In- formation Criterion (DIC), similar to Celeux et al. (2006), but which is suitable for use with generalized linear models (GLMs) when covariates are missing at random. We define three different DICs: the marginal, where the missing data are averaged out of the likelihood; the complete, where the joint likelihood for response and covariates is considered; and the naive, where the likelihood is found assuming the missing values are parameters. These three versions have different computational complexities. We investigate through simulation the performance of these three different DICs for GLMs consisting of normally, binomially and multinomially distributed data with missing covariates having a normal distribution. We find that the marginal DIC and the estimate of the effective number of parameters, pD, have desirable properties appropriately indicating the true model for the response under differing amounts of missingness of the covariates. We find that the complete DIC is inappropriate generally in this context as it is extremely sensitive to the degree of missingness of the covariate model. Our new methodology is illustrated by analysing the results of a community survey.

Diss. (sammanfattning) Uppsala : Sveriges lantbruksuniviversitet, 2006. / Härtill 7 uppsatser.

Thesis (Ph.D.)--Mississippi State University. Department of Mathematics and Statistics. / Title from title screen. Includes bibliographical references.

Thesis (M.Sc. (Statistics)) - Rhodes University, 2007. / A thesis submitted in partial fulfilment of the requirements for the degree of Master of Science.

abstract: Yield is a key process performance characteristic in the capital-intensive semiconductor fabrication process. In an industry where machines cost millions of dollars and cycle times are a number of months, predicting and optimizing yield are critical to process improvement, customer satisfaction, and financial success. Semiconductor yield modeling is essential to identifying processing issues, improving quality, and meeting customer demand in the industry. However, the complicated fabrication process, the massive amount of data collected, and the number of models available make yield modeling a complex and challenging task. This work presents modeling strategies to forecast yield using generalized linear models (GLMs) based on defect metrology data. The research is divided into three main parts. First, the data integration and aggregation necessary for model building are described, and GLMs are constructed for yield forecasting. This technique yields results at both the die and the wafer levels, outperforms existing models found in the literature based on prediction errors, and identifies significant factors that can drive process improvement. This method also allows the nested structure of the process to be considered in the model, improving predictive capabilities and violating fewer assumptions. To account for the random sampling typically used in fabrication, the work is extended by using generalized linear mixed models (GLMMs) and a larger dataset to show the differences between batch-specific and population-averaged models in this application and how they compare to GLMs. These results show some additional improvements in forecasting abilities under certain conditions and show the differences between the significant effects identified in the GLM and GLMM models. The effects of link functions and sample size are also examined at the die and wafer levels. The third part of this research describes a methodology for integrating classification and regression trees (CART) with GLMs. This technique uses the terminal nodes identified in the classification tree to add predictors to a GLM. This method enables the model to consider important interaction terms in a simpler way than with the GLM alone, and provides valuable insight into the fabrication process through the combination of the tree structure and the statistical analysis of the GLM. / Dissertation/Thesis / Ph.D. Industrial Engineering 2011

Pour les structures en bois comme pour les autres techniques de construction, les planchers jouent un rôle important dans le comportement global d’un bâtiment soumis à une sollicitation sismique. En plus de la reprise des efforts verticaux, ces planchers assurent le transfert des efforts horizontaux vers les murs de contreventement (fonction diaphragme). Pour cette fonction, le plancher peut être assimilé à une poutre courte disposée horizontalement et supportée par plusieurs appuis semi-rigides (murs de contreventement). En conséquence, la distribution des efforts sur les murs de contreventement est directement liée au rapport entre leurs raideurs et celle du plancher dans son plan.Les planchers en bois traditionnels sont généralement composés par une poutraison (des solives, des entretoises et des poutres de chaînage) et une surface plane composée d’une juxtaposition de panneaux. La connexion panneaux-poutraison est généralement réalisée par des clous. Le dimensionnement de ces planchers est souvent piloté par les charges verticales et les caractéristiques mécaniques dans le plan sont rarement contrôlées.Le comportement mécanique des planchers diaphragmes en bois est peu étudié dans la littérature. La présente thèse aborde ce comportement en s’appuyant sur trois approches : analytique, numérique (M.E.F.) et expérimentale. L’approche analytique s’inspire de celle utilisée par les codes de calcul sismique nord-américains. La modélisation numérique prend en compte la configuration géométrique réelle des diaphragmes ainsi que les effets non-linéaires matériels et géométriques. Deux types de modèles numériques sont proposés : « détaillé » et « simplifié ». Le modèle détaillé, utilisant des éléments de poutres et de coques, permet d’étudier la distribution des contraintes internes au sein du plancher. Le modèle simplifié, moins coûteux, s’appuie sur des éléments poutres et barres. Il permet de réaliser des analyses de sensibilité et il peut être utilisé dans des analyses de structures avec géométries complexes.Une campagne expérimentale est conduite sur des planchers en bois. Deux planchers (2,4x7,2m2), avec et sans entretoises, sont testés en flexion dans leur plan. Une instrumentation riche composée de capteurs de déplacement, d’inclinomètres et d’un système de mesure sans contact, sont utilisés pour suivre le comportement de ces planchers aux niveaux global et local. Les observations expérimentales montrent un comportement fortement non-linéaire avec une influence importante des entretoises sur la raideur et la résistance. La rupture des panneaux est due à la flexion à mi portée ou à la compression près des angles en contact. La comparaison avec les résultats expérimentaux a permis de valider les modèles numériques développés. Ces modèles révèlent une forte participation de la connexion panneaux-poutraison aux comportements mécaniques des planchers aux niveaux global et local.Le modèle numérique simplifié est utilisé pour mener une étude de sensibilité des planchers avec trémies. Ainsi, treize configurations de planchers sont analysées en vue d’identifier les positionnements les plus défavorables des trémies. Par ailleurs, l’application du modèle analytique montre qu’il représente correctement la phase linéaire du comportement des planchers qui est un paramètre utilisable dans les analyses à l’état limite de service. Les résultats numériques et expérimentaux obtenus constituent une bonne base pour étendre le domaine d’application du modèle analytique. / In timber structures as other, floors play an important role in the overall behaviour of a building subjected to seismic loading. In addition to the vertical forces, these floors transfer horizontal forces to the shear walls (diaphragm function). For this function, the floor can be considered as a short beam placed horizontally and supported by semi-rigid supports (shear walls). Therefore, the force distribution on the shear walls is directly related to the in-plane stiffness ratio between floors and shear walls.Traditional wooden floors are usually composed by joists framing (joist, struts, chords beams) and a surface made by juxtaposed panels. The panel-joist framing connection is usually made by nails. Size of the floors elements is often driven by the vertical loads, thus mechanical properties in the plane are rarely checked.The in-plane behaviour of timber floor is poorly studied in the literature. This thesis studies this behaviour using three approaches: analytical, numerical (FEM) and experimental. The analytical approach is based on that used in North American codes of seismic design. Numerical models take into account the actual geometry of the timber diaphragms and nonlinear effects (material and geometric). Two types of numerical models are proposed: "detailed" and "simplified". The detailed model, based on beams and shell elements, allows studying the distribution of internal stresses in the floor. The simplified model uses beams and bar elements. It allows carrying out sensitivity analysis and can be used to study complex geometries of floors.An experimental campaign is conducted on timber floors. Two floors (2,4 x7,2m2), in blocked or unblocked configuration, are tested in bending in their plane. Displacement transducers, inclinometers and a camera measuring system are used for monitoring the behaviour of floors at global and local levels. The experimental observations show a highly nonlinear behaviour with a significant influence of struts on the stiffness and strength. Panels failure due to bending and compression in the corners were observed. The comparison between experimental and numerical results allows validating the numerical models. These models show a significant participation of the panels-joist framing connection on the global and local mechanical behaviour of floors.The simplified model is used to conduct a sensitivity analysis of floors with hoppers. Thirteen floors configurations are analyzed to identify the most unfavourable configurations of hoppers position. Furthermore, the analytical model represents well the linear phase of floors behaviour. The numerical and experimental results obtained provide a good basis to expand the scope of the analytical model. / Tanto en las estructuras de madera como en otras técnicas de construcción, las losas juegan un importante rol en el comportamiento global de un edificio sometido a una solicitación sísmica. Además de las cargas verticales, las losas deben transmitir las cargas horizontales hacia los muros de corte (función diafragma). En esta función, la losa puede ser tratada como una viga corta puesta horizontalmente y apoyada sobre soportes semi-rígidos (muros de corte). En consecuencia, la distribución de fuerzas entre los distintos muros de corte es determinada por la relación de rigidez entre estos muros y la losa.Las losas tradicionales de madera son generalmente compuestas por un envigado (vigas transversales, canes o separadores, vigas periféricas) y una superficie construida con paneles yuxtapuestos. La conexión entre los paneles y el envigado es generalmente realizada con clavos. El dimensionamiento de estos elementos es habitualmente determinado por las cargas verticales, de esta manera las propiedades mecánicas en el plano de estas losas, son raramente controladas.El comportamiento mecánico de los diafragmas de madera ha sido poco estudiado en la literatura. Esta tesis aborda este comportamiento a partir de tres enfoques: analítico, numérico (Elementos Finitos) y experimental. El método analítico considerado se basa en las ecuaciones propuestas por los códigos de diseño sísmico norteamericanos. La modelización numérica tiene en cuenta la configuración geométrica real de las losas y los efectos no-lineares de origen material y geométrico. Dos tipos de modelos numéricos son propuestos: "detallado" y "simplificado". El modelo detallado, es construido a partir de elementos finitos de tipo vigas y planos. Este modelo es utilizado principalmente para estudiar la distribución de esfuerzos internos en las losas. El modelo simplificado, construido a partir de elementos de vigas y barras, es utilizado para conducir análisis de sensibilidad.Un estudio experimental sobre losas de madera es presentado. Dos losas (2,4 x7, 2m2) con y sin separadores son testeadas en flexión en su plano. Un sistema de instrumentación compuesto por transductores de desplazamiento, inclinómetros y videocámaras, fueron utilizados para monitorear el comportamiento global y local de las losas. Las observaciones experimentales muestran un comportamiento altamente no-lineal con una fuerte influencia de los separadores sobre la rigidez y resistencia de la estructura. Los modos de ruptura observados corresponden a la ruptura de paneles en flexión y a la compresión local debido a transmisión de esfuerzos por contacto. La comparación de resultados experimentales y numéricos permite validar los modelos propuestos. Estos modelos muestran una fuerte participación de la conexión paneles-envigado sobre el comportamiento mecánico global y local de las losas.El modelo numérico simplificado es utilizado para llevar a cabo un análisis de sensibilidad de losas con aberturas de escalera. Trece configuraciones de losas son analizadas para identificar la posición de abertura más desfavorable. Por otra parte, el modelo analítico representa correctamente el comportamiento de las losas en fase lineal. Los resultados numéricos y experimentales obtenidos proporcionan una buena base para ampliar el dominio de aplicación de este modelo analítico.

A presente dissertação aborda a otimização de processos industriais através da utilização de projeto de experimentos. Em experimentos planejados, variáveis de respostas são usualmente consideradas como normalmente distribuídas. No entanto, em algumas situações, tal suposição é violada; por exemplo, quando respostas expressam contagens, podendo assumir somente valores inteiros e não-negativos. Nesses casos, é mais provável que as respostas sigam uma distribuição de Poisson e, em sua modelagem, deve-se utilizar os modelos lineares generalizados (MLG), adequados para respostas da chamada família exponencial, à qual pertence a distribuição de Poisson. Se ainda persiste a dúvida quanto ao comportamento das respostas, o modelo de quase-verossimilhança também é uma alternativa possível. Esta dissertação apresenta a reanálise de um experimento, apresentado em Arriba (2005), onde algumas respostas são categóricas. Na análise original do experimento, respostas categóricas foram modeladas através de regressão dos mínimos quadrados ordinários, desconsiderando a violação do pressuposto de normalidade das respostas. Na reanálise aqui apresentada, as variáveis são corretamente abordadas usando-se os modelos lineares generalizados. Como o objetivo do trabalho de Arriba (2005) era a otimização de um processo descrito por múltiplas respostas, comparam-se os resultados da otimização mediante as duas estratégias de modelagem das respostas, além de se propor um método alternativo, mais simplificado, de otimização experimental. / This dissertation deals with the optimization of industrial processes using Design of Experiments. In designed experiments, response variables are often considered as normally distributed. However, in some situations, such assumption is violated; for example, when responses express counts, and only non-negative integers numbers may come up as outcomes. In these cases, it is likely that responses follow a Poisson distribution which is then modeled by generalized linear models (GLM), since such distribution is a member of the exponential family. If a question still holds on the responses behavior, their modeling through the quasi-likelihood method is another option that should be considered. This dissertation analyses an experiment performed by Arriba (2005), where some responses are of categorical nature. In the original analysis presented by Arriba (2005), categorical responses were modeled using ordinary least squares regression, violating the normality assumption associated with that method. In the analysis presented here, variables are appropriately modeled using GLM. Since the objective in Arriba (2005) was optimizing a multiresponse process, results from the two optimization processes are compared. In addition, an alternative experimental optimization method, simpler than the one in Arriba (2005), is also presented.

abstract: City administrators and real-estate developers have been setting up rather aggressive energy efficiency targets. This, in turn, has led the building science research groups across the globe to focus on urban scale building performance studies and level of abstraction associated with the simulations of the same. The increasing maturity of the stakeholders towards energy efficiency and creating comfortable working environment has led researchers to develop methodologies and tools for addressing the policy driven interventions whether it’s urban level energy systems, buildings’ operational optimization or retrofit guidelines. Typically, these large-scale simulations are carried out by grouping buildings based on their design similarities i.e. standardization of the buildings. Such an approach does not necessarily lead to potential working inputs which can make decision-making effective. To address this, a novel approach is proposed in the present study. The principle objective of this study is to propose, to define and evaluate the methodology to utilize machine learning algorithms in defining representative building archetypes for the Stock-level Building Energy Modeling (SBEM) which are based on operational parameter database. The study uses “Phoenix- climate” based CBECS-2012 survey microdata for analysis and validation. Using the database, parameter correlations are studied to understand the relation between input parameters and the energy performance. Contrary to precedence, the study establishes that the energy performance is better explained by the non-linear models. The non-linear behavior is explained by advanced learning algorithms. Based on these algorithms, the buildings at study are grouped into meaningful clusters. The cluster “mediod” (statistically the centroid, meaning building that can be represented as the centroid of the cluster) are established statistically to identify the level of abstraction that is acceptable for the whole building energy simulations and post that the retrofit decision-making. Further, the methodology is validated by conducting Monte-Carlo simulations on 13 key input simulation parameters. The sensitivity analysis of these 13 parameters is utilized to identify the optimum retrofits. From the sample analysis, the envelope parameters are found to be more sensitive towards the EUI of the building and thus retrofit packages should also be directed to maximize the energy usage reduction. / Dissertation/Thesis / Masters Thesis Architecture 2017

Submitted by Johnny Rodrigues (johnnyrodrigues@ufcg.edu.br) on 2018-07-16T19:54:29Z No. of bitstreams: 1 ARELI MESQUITA DA SILVA - DISSERTAÇÃO PPGMAT 2007..pdf: 701919 bytes, checksum: 78ea7b65513f1fe6d83acdb4f3030b43 (MD5) / Made available in DSpace on 2018-07-16T19:54:29Z (GMT). No. of bitstreams: 1 ARELI MESQUITA DA SILVA - DISSERTAÇÃO PPGMAT 2007..pdf: 701919 bytes, checksum: 78ea7b65513f1fe6d83acdb4f3030b43 (MD5) Previous issue date: 2007-07 / Séries temporais envolvendo dados angulares aparecem nas mais diversas áreas do conhecimento. Por exemplo, na perfuração de um poço petrolífero direcional, o deslocamento da broca de perfuração, ao longo da trajetória do poço, pode ser considerado uma realização de uma série temporal de dados angulares. Um dos interesses, neste contexto, consiste em realizar previsões de posicionamentos futuros da broca de perfuração, as quais darão mais apoio ao engenheiro de petróleo na tomada de decisão de quando e como interferir na trajetória de um poço, de modo que este siga o curso planejado. Neste trabalho, estudamos algumas classes de modelos que podem ser utilizados para a modelagem desse tipo de série. / Time series involving angular data appear in many diverse areas of scientific knowledge. For example, in the drilling of a directional oil well, the displacement of the drill, along the path of the well, can be considered as an angular data time series. One of the objectives, in this context, consists in carrying out forecasts of the future positions of the drill, which will give more support to the petroleum engineer in the decision-making of when and how interfere in the path of a well, so that this follows the planned course. In this work, we study some classes of models that can be utilized for the modeling of that kind of series.