Bayesian subset selection suffers from three important difficulties: assigning priors over model space, assigning priors to all components of the regression coefficients vector given a specific model and Bayesian computational efficiency (Chen et al., 1999). These difficulties become more challenging in Bayesian quantile regression framework when one is interested in assigning priors that depend on different quantile levels. The objective of Bayesian quantile regression (BQR), which is a newly proposed tool, is to deal with unknown parameters and model uncertainty in quantile regression (QR). However, Bayesian subset selection in quantile regression models is usually a difficult issue due to the computational challenges and nonavailability of conjugate prior distributions that are dependent on the quantile level. These challenges are rarely addressed via either penalised likelihood function or stochastic search variable selection (SSVS). These methods typically use symmetric prior distributions for regression coefficients, such as the Gaussian and Laplace, which may be suitable for median regression. However, an extreme quantile regression should have different regression coefficients from the median regression, and thus the priors for quantile regression coefficients should depend on quantiles. This thesis focuses on three challenges: assigning standard quantile dependent prior distributions for the regression coefficients, assigning suitable quantile dependent priors over model space and achieving computational efficiency. The first of these challenges is studied in Chapter 2 in which a quantile dependent prior elicitation scheme is developed. In particular, an extension of the Zellners prior which allows for a conditional conjugate prior and quantile dependent prior on Bayesian quantile regression is proposed. The prior is generalised in Chapter 3 by introducing a ridge parameter to address important challenges that may arise in some applications, such as multicollinearity and overfitting problems. The proposed prior is also used in Chapter 4 for subset selection of the fixed and random coefficients in a linear mixedeffects QR model. In Chapter 5 we specify normal-exponential prior distributions for the regression coefficients which can provide adaptive shrinkage and represent an alternative model to the Bayesian Lasso quantile regression model. For the second challenge, we assign a quantile dependent prior over model space in Chapter 2. The prior is based on the percentage bend correlation which depends on the quantile level. This prior is novel and is used in Bayesian regression for the first time. For the third challenge of computational efficiency, Gibbs samplers are derived and setup to facilitate the computation of the proposed methods. In addition to the three major aforementioned challenges this thesis also addresses other important issues such as the regularisation in quantile regression and selecting both random and fixed effects in mixed quantile regression models.

Recent advances in genome editing and metabolic engineering enabled a precise construction of de novo biosynthesis pathways for high-value natural products. One important design decision to make for the engineering of heterologous biosynthesis systems is concerned with which foreign metabolic genes to introduce into a given host organism. Although this decision must be made based on multifaceted factors, a major one is the suitability of pathways for the endogenous metabolism of a host organism, in part because the efficacy of heterologous biosynthesis is affected by competing endogenous pathways. To address this point, we developed an open-access web server called MRE (metabolic route explorer) that systematically searches for promising heterologous pathways by considering competing endogenous reactions in a given host organism. MRE utilizes reaction Gibbs free energy information. However, 25% of the reactions do not have accurate estimations or cannot be estimated. To address this issue, we developed a method called FC (fingerprint contribution) to provide a more accurate and complete estimation of the reaction free energy. To rationally design a productive heterologous biosynthesis system, it is essential to consider the suitability of foreign reactions for the specific endogenous metabolic infrastructure of a host. For a given pair of starting and desired compounds in a given chassis organism, MRE ranks biosynthesis routes from the perspective of the integration of new reactions into the endogenous metabolic system. For each promising heterologous biosynthesis pathway, MRE suggests actual enzymes for foreign metabolic reactions and generates information on competing endogenous reactions for the consumption of metabolites. The URL of MRE is http://www.cbrc.kaust.edu.sa/mre/. Accurate and wide-ranging prediction of thermodynamic parameters for biochemical reactions can facilitate deeper insights into the workings and the design of metabolic systems. Here, we introduce a machine learning method, referred to as fingerprint contribution (FC), with chemical fingerprint-based features for the prediction of the Gibbs free energy of biochemical reactions. From a large pool of 2D fingerprint-based features, this method systematically selects a small number of relevant ones and uses them to construct a regularized linear model. FC is freely available for download at http://sfb.kaust.edu.sa/Pages/Software.aspx.

Time-synchronization in wireless ad-hoc sensor networks is a crucial piece of infrastructure. Thus, it is a fundamental design problem to have a good clock syn- chronization amongst the nodes of wireless ad-hoc sensor networks. Motivated by this fact, in this thesis, the joint maximum likelihood (JML) estimator for relative clock phase offset and skew under the exponential noise model for the reference broadcast synchronization protocol is formulated and found via a direct algorithm. The Gibbs Sampler is also proposed for joint estimation of relative clock phase offset and skew, and shown to provide superior performance compared to the JML-estimator. Lower and upper bounds for the mean-square errors (MSE) of the JML-estimator and the Gibbs Sampler are introduced in terms of the MSE of the uniform minimum variance unbiased estimator and the conventional best linear unbiased estimator, respectively. The suitability of the Gibbs Sampler for estimating additional unknown parameters is shown by applying it to the problem in which synchronization of clock drift is also needed.

Thesis (Ph. D.)--West Virginia University, 2002. / Title from document title page. Document formatted into pages; contains xi, 107 p. : ill. (some col.). Includes abstract. Includes bibliographical references (p. 102-107).

The mechanical behavior of asphalt binders is nonlinear. The binders exhibit shear thinning/thickening behavior in steady shear tests and non-proportational behavior in other standard viscoelastic tests such as creep-recovery or stress relaxation tests. Moreover, they develop normal stress differences even in simple shear flows - a characteristic feature of nonlinear viscoelastic behavior. Many researchers have asserted the importance of considering the nonlinearity of the mechanical behavior of asphalt binders for accurately estimating their performance under field conditions, and for comparing and ranking them accordingly. In order to do so, it is necessary to have a robust and reliable nonlinear viscoelastic model. However, most of the models available in the literature do not capture the various features of the nonlinear response of asphalt binders accurately. Those that could are too complicated and still possess other shortcomings. Considering these issues, a new nonlinear viscoelastic model is developed here using a new Gibbs-potential based thermodynamic framework. The model is then corraborated with data from experiments in which the shear-thinning behavior and the nonproportional creep-recovery behavior were observed together. Finally, the model is used to evaluate the various criteria available for predicting rutting susceptibility of asphalt binders. Results of the analysis of the rutting prediction criteria show that each criterion characterizes the resistance to permanent strain shown by asphalt binders over a different range of applied stress - the zero-shear viscosity at very low stress levels, the Superpave criterion at very high stress levels and the MSCR test in the intermediate range of stresses.

Techn. University, Diss., 2005--Darmstadt.

Thesis (Ph. D.)--University of Washington, 2000. / Vita. Includes bibliographical references (leaves 105-110).

Item response theory (IRT) is a newer and improved theory compared to the classical measurement theory. The fully Bayesian approach shows promise for IRT models. However, it is computationally expensive, and therefore is limited in various applications. It is important to seek ways to reduce the execution time and a suitable solution is the use of high performance computing (HPC). HPC offers considerably high computational power and can handle applications with high computation and memory requirements. In this work, we have modified the existing fully Bayesian algorithm for 2PNO IRT models so that it can be run on a high performance parallel machine. With this parallel version of the algorithm, the empirical results show that a speedup was achieved and the execution time was reduced considerably.

Item response theory (IRT) is a newer and improved theory compared to the classical measurement theory. The fully Bayesian approach shows promise for IRT models. However, it is computationally expensive, and therefore is limited in various applications. It is important to seek ways to reduce the execution time and a suitable solution is the use of high performance computing (HPC). HPC offers considerably high computational power and can handle applications with high computation and memory requirements. In this work, we have applied two different parallelism methods to the existing fully Bayesian algorithm for 2PNO IRT models so that it can be run on a high performance parallel machine with less communication load. With our parallel version of the algorithm, the empirical results show that a speedup was achieved and the execution time was considerably reduced.

Made available in DSpace on 2014-06-11T19:26:07Z (GMT). No. of bitstreams: 0 Previous issue date: 2012-05-25Bitstream added on 2014-06-13T19:54:01Z : No. of bitstreams: 1 gama_mpm_me_jabo.pdf: 190063 bytes, checksum: b6fc2664d1583f57dbb78c9f59a29b61 (MD5) / Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / O objetivo desse trabalho foi estimar parâmetros genéticos para o caráter tempo e colocação final em corridas de cavalos Puro Sangue Inglês (PSI) com procedimentos Bayesiano e Thurstoniano, a fim de fornecer subsídios para a seleção de reprodutores e consequente melhoramento genético da raça no Brasil. A partir de dados fornecidos pela empresa Turf Total Ltda foram consideradas 251.754 informações de tempo e 272.277 informações de colocações finais em 34.316 corridas de cavalos PSI ocorridas entre 1992 e 2011 em 6 hipódromos do país, para as distâncias de 1.000, 1.300, 1.600 e 2.000 metros. Os efeitos considerados fixos foram idade, sexo, posição de largada e páreo para as análises de tempo, e sexo, idade, posição de largada, páreo e nível de dificuldade da corrida para as análises de colocações finais. As herdabilidades e correlações genéticas para tempo foram estimadas utilizando inferências bayesianas, ao passo que para as estimativas de herdabilidade de colocação final utilizou-se o modelo Thurstoniano. As estimativas de herdabilidade para tempo em análises unicaraterísticas. foram semelhantes às encontradas na literatura, e variaram entre 0,31 e 0,04 e as repetibilidades encontradas variaram de 0,61 a 0,22, respectivamente com o aumento das distâncias. As estimadas para colocação final variaram de 0,57 a 0,21, apresentando a mesma tendência que as herdabilidades para tempo. As estimativas de herdabilidade para tempo na análise multicaracterística variaram entre 0,34 e 0,15 com repetibilidade entre 0,63 e 0,36. Nessa análise, a seleção para tempo também mostrou-se mais eficiente em distâncias menores, onde as herdabilidades foram maiores. As estimativas de correlações genéticas foram positivas e variaram de 0,47 a 0,97. Conclui-se que, em distâncias mais curtas, a seleção tanto para tempo... / The objective of this study was to estimate genetic paremeters for racing time and final rank in Thoroughbred horses using Bayesian and Thurstonian procedures, in order to provide data that contribute for selection and the consequent genetic improvement of the breed in Brazil. Data were provided by the company Turf Total Ltda. and consisted of 251,754 racing time records and 272,277 final rank records obtained from 34,316 Thoroughbred races (distances of 1,000, 1,300, 1,600 and 2,000 m) that occurred between 1992 and 2011 on six race tracks. Fixed effects included age, sex, post-position and race for the analysis of racing time, and sex, age, post-position, race and level of difficulty for final rank analysis. The heritabilities for racing time and final rank and the genetic correlations between racing times were estimated by Bayesian inference. In addition, a Thurstonian model was used to estimate the heritability for final rank. The heritability estimates for racing time in one-trait analysis were similar to those reported in the literature and ranged from 0.31 to 0.04. Repeatability estimates tended to decrease with increasing race distance (0.61 to 0.22) .The heritabilities estimates for final rank ranged from 0.57 to 0.21 and showed the same trend as the heritabilities for time. The heritability estimates for racing time obtained by multi-trait analysis ranged form 0.34 to 0.15, with repeatabilities of 0.63 to 0.36 at the distances studied. Multi-trait analysis also showed that selection for racing time was more efficient at shorter distances when heritabilities were higher. The genetic correlations were all positive and ranged from 0.47 to 0.97. In conclusion, selection for both racing time and final rank is more efficient at shorter distances... (Complete abstract click electronic access below)