L'objectif de cette thèse est de fournir des outils d'aide à la décision pour piloter les plateformes logistiques à court de moyen terme. La première partie décrit la problématique concernée et les notions essentielles dans le cadre des chaînes logistiques. Dans la deuxième partie, le problème de la planification est étudié, nous proposons des modèles linéaires pour minimiser les coûts de personnel, qui prennent en compte les flux : leurs variations saisonnières, la possibilité de les négocier localement en amont et en aval, ainsi que leur organisation et celle du travail. Ainsi, l'outil peut être utilisé dans la coordination des flux entres les partenaires de la chaîne livrées en amont et en aval de la plateforme et la négociation des dates de livraison. Ces modèles sont testés et validés sur des instances générées aléatoirement, sur des configurations inspirées de problèmes réels. Dans la troisième partie, nous travaillons sur l'ordonnancement des activités de préparation de commandes. Ici, nous combinons deux familles de contraintes difficiles : l'arrivée de composants (ressources consommables) à des dates et en quantités connues à l'amont de la plateforme, et des tournées de livraison à dates fixées à l'aval. Trois cas particuliers sont étudiés, selon la façon dont les tournées sont organisées. Nous proposons des procédures par séparation et évaluation pour ces problèmes, et un modèle linéaire en nombres entiers pour le cas le plus simple. Des expériences sont faites sur des familles d'instances générées aléatoirement et de manière partiellement hétérogène. Plusieurs perspectives de généralisation sont proposées / The aim of this thesis is to provide decision support systems to control logistic platforms at the mid-term and short-term levels. Several problems and main notions concerning logistic platform context are described in the first part. In the second part, planning problems are studied. Two linear programming models are proposed to minimize the workforce costs. These models take into account several characteristics : seasonal flow variations, work and flow organization in the platform, and local negotiations of the upstream and downstream flows. In consequence, our decision support system can be used in the flow coordination between supply chain partners. Two types of negotiations are considered : negotiations of upstream and downstream delivered quantities and negotiation of delivery dates. These models have been tested on pertinent randomly generated instances inspired from concerete problems. In the third part of the thesis, the external flows of the platforme are assumed to be fixed. Orders preparation scheduling problem inside the platform is considered. Two families of strong contraints are combined : staircase availability of components (consumable resources) and dixed delivery dates. According to the way the downstream deliveries are organized and penalised, three different cases (based on industrial applications) have been studied. We proposed three branch and bound procedures for these problems, and an integer linear program for the easiest problem. Experimental analysis has been done over heterogeneous randomly generated instance families. In the last part, a series of perspectives for this work are proposed

In recent years, statistical analyses, algorithms, and modeling of big data have been constrained due to computational complexity. Further, the added complexity of relationships among response and explanatory variables, such as higher-order interaction effects, make identifying predictors using standard statistical techniques difficult. These difficulties are only exacerbated in the case of small sample sizes in some studies. Recent analyses have targeted the identification of interaction effects in big data, but the development of methods to identify higher-order interaction effects has been limited by computational concerns. One recently studied method is the Feasible Solutions Algorithm (FSA), a fast, flexible method that aims to find a set of statistically optimal models via a stochastic search algorithm. Although FSA has shown promise, its current limits include that the user must choose the number of times to run the algorithm. Here, statistical guidance is provided for this number iterations by deriving a lower bound on the probability of obtaining the statistically optimal model in a number of iterations of FSA. Moreover, logistic regression is severely limited when two predictors can perfectly separate the two outcomes. In the case of small sample sizes, this occurs quite often by chance, especially in the case of a large number of predictors. Bhattacharyya distance is proposed as an alternative method to address this limitation. However, little is known about the theoretical properties or distribution of B-distance. Thus, properties and the distribution of this distance measure are derived here. A hypothesis test and confidence interval are developed and tested on both simulated and real data.

Organizations around the world have attempted to implement the concepts of the Toyota Production System (TPS), commonly referred to as Lean, with limited sustainable success. The central principles of TPS, continuous improvement and respect for people, are grounded in the Japanese values of Monozukuri and Hitozukuri. Monozukuri deals with creating or making a product, while Hitozukuri conveys the idea of developing people through learning. In order for organizations to adopt these values they must have a system that engages employees at all levels in applying problem solving to improve their work. This research uses organizational assessments obtained from a variety of organizations implementing the lean approach using the Monozukuri and Hitozukuri values, referred to as the True Lean System (TLS). This research uses an inductive research approach to identify and analyze factors that impact the use of problem solving within organizations implementing a TLS. First, the qualitative assessment data is studied using textual analysis to identify themes impacting TLS. This analysis identified three topics as the highest weighted themes: number of problem solving methods, standardization, and employee roles. This qualitative data is then transformed using an integrated design model to systematically code the information into quantitative numerical data. Finally, this data was analyzed statistically by logistic regression to identify the factors impacting the use of problem solving within these organizations. The results from the logistic regression suggest that the most successful problem solving organizations have established standards for work and training employees; as well as, a single problem solving method that all employees use when identifying and implementing continuous improvement ideas. Which leads to the conclusion, in order for an organization to sustain the concepts of TPS, there must be a focus on defining clear standardized work, training, and the implementation of a single problem solving method.

The purpose of this research is to investigate the performance of some ridge regression estimators for the logistic regression model in the presence of moderate to high correlation among the explanatory variables. As a performance criterion, we use the mean square error (MSE), the mean absolute percentage error (MAPE), the magnitude of bias, and the percentage of times the ridge regression estimator produces a higher MSE than the maximum likelihood estimator. A Monto Carlo simulation study has been executed to compare the performance of the ridge regression estimators under different experimental conditions. The degree of correlation, sample size, number of independent variables, and log odds ratio has been varied in the design of experiment. Simulation results show that under certain conditions, the ridge regression estimators outperform the maximum likelihood estimator. Moreover, an empirical data analysis supports the main findings of this study. This thesis proposed and recommended some good ridge regression estimators of the logistic regression model for the practitioners in the field of health, physical and social sciences.

Occupational and environmental epidemiological studies often involve ordinal data, including antibody titer data, indicators of health perceptions, and certain psychometrics. Ideally, such data should be analyzed using approaches that exploit the ordinal nature of the scale, while making a minimum of assumptions. In this work, we first review and illustrate the analytical technique of ordinal logistic regression called the "proportional odds model". This model, which is based on a constrained ordinal model, is considered the most popular ordinal model. We use hypothetical data to illustrate a situation where the proportional odds model holds exactly, and we demonstrate through derivations and simulations how using this model has better statistical power than simple logistic regression. The section concludes with an example illustrating the use of the model in avian and swine influenza research. In the middle section of this work, we show how the proportional model assumption can be relaxed to a less restrictive model called the "trend odds model". We demonstrate how this model is related to latent logistic, normal, and exponential distributions. In particular, scale changes in these potential latent distributions are found to be consistent with the trend odds assumption, with the logistic and exponential distributions having odds that increase in a linear or nearly linear fashion. Actual data of antibody titer against avian and swine influenza among occupationally- exposed participants and non-exposed controls illustrate the fit and interpretation of the proportional odds model and the trend odds model. Finally, we show how to perform a multivariable analysis in which some of the variables meet the proportional model assumption and some meet the trend odds assumption. Likert-scaled data pertaining to violence among middle school students illustrate the fit and interpretation of the multivariable proportional-trend odds model. In conclusion, the proportional odds model provides superior power compared to models that employ arbitrary dichotomization of ordinal data. In addition, the added complexity of the trend odds model provides improved power over the proportional odds model when there are moderate to severe departures from proportionality. The increase in power is of great public health relevance in a time of increasingly scarce resources for occupational and environmental health research. The trend odds model indicates and tests the presence of a trend in odds, providing a new dimension to risk factors and disease etiology analyses. In addition to applications demonstrated in this work, other research areas in occupational and environmental health can benefit from the use of these methods. For example, worker fatigue is often self-reported using ordinal scales, and traumatic brain injury recovery is measured using recovery scores such as the Glasgow Outcome Scale (GOS).

When the delicate hormonal balance in early pregnancy is disrupted, the consequences can be significant. We have a poor understanding of the "cross-talk" in the fetal/placental/ovarian axis that occurs throughout pregnancy and is essential for normal fetal development. This lack of knowledge challenges our ability to recognize disruptions in this axis that may be a signal for future disease. As a result, our ability to apply preventive measures against adverse obstetric outcomes, such as preterm birth (PTB), are quite limited. Attempts to predict PTB using biomarkers of feto-placental health have been largely unsuccessful, but no one has considered the inclusion of ovarian biomarkers in these models. Anti-Mullerian hormone (AMH) is a biomarker of ovarian activity that has recently been found to decline in early pregnancy at a time that corresponds to the involution of the corpus luteum (CL). The signal for CL involution is believed to originate from the placenta; therefore, the AMH levels in pregnancy may reflect the degree of ovarian up or down-regulation based on feto-placental needs. As the major function of the CL in pregnancy is the production of progesterone, which acts as an anti-inflammatory agent in the placental bed, changes in CL-derived progesterone could result in higher or lower degrees of placental inflammation. Therefore, monitoring the changes in AMH levels may provide insight into the inflammatory state of the placenta which could then be used as a signal for possible adverse obstetric outcomes resulting from a pro-inflammatory state, such as PTB. The first aim of this project was to test the hypothesis of an association between AMH levels in early pregnancy and PTB risk. When the differences in AMH levels between the 1st and 2nd trimesters of pregnancy were stratified by the level of maternal serum alpha-fetoprotein (MSAFP) and controlled for maternal weight gain between trimesters, small or absent decreases in AMH levels were associated with a higher probability of preterm birth. However, when AMH was modeled alone, no significant associations were found. The need for changes in multiple biomarkers in the fetal/placental/ovarian axis suggests that a change is only significant if it can impact multiple axis points. Therefore, models that included two biomarkers from different part of the axis would find stronger associations than two biomarkers from a single point (e.g. two feto-placental biomarkers), and monitoring these changes may help identify women at risk for PTB. The strategy of the second aim was to determine if the changes in AMH levels in early pregnancy could be used to predict time to delivery. Again, only when the risks of AMH and MSAFP were combined was a significant, dose-dependent relationship found with time to delivery. In women with an MSAFP of >1 multiple of the median (MoM), smaller declines and/or elevations in AMH levels were significantly associated with shorter times to delivery. In fact, 19% of women in the highest risk group delivered prior to 32 weeks gestation compared to 7% in the lowest risk group, and all infants who delivered prior to 24 weeks gestation were in the highest risk category. Thus, the amount of change in the AMH level when MSAFP is elevated may reflect the level of disruption in the fetal/placental/ovarian axis, which can then be used to predict time to delivery. Finally, the third aim of this study was to determine if AMH levels were associated with a pro-inflammatory placental state other than PTB. The degree of placental inflammation is known to vary by fetal gender, with male placentas having higher levels of inflammation compared to female placentas. When AMH levels were compared between women with male vs. female fetuses in early pregnancy, 1st trimester AMH levels were found to be lower when carrying a male fetus. Further, sexually-dimorphic patterns in AMH levels were seen between genders when stratified by birth outcome (term vs. preterm delivery). The stronger ovarian response seen in women with female fetuses suggests a better survival function and may account for the discrepancies between PTB rates in males and females. This also strengthens our hypothesis that the dynamic changes in AMH levels reflect the degree of placental inflammation and the need for CL-derived progesterone. This project demonstrates that the changes in AMH levels may be representative of the cross-talk occurring in the fetal/placental/ovarian axis in early pregnancy. Further, changes in AMH levels may be an indication of the amount of inflammation in the placenta and the physiologic need for higher levels of progesterone to control this inflammatory state when considered along with MSAFP. Therefore, the consideration of AMH levels as a biomarker of ovarian activity along with biomarkers of feto-placental health may provide clinically useful information about the development of future diseases such as preterm birth.

Bundling is pervasive in the market, examples include desktop computer bundles, digital single-lens reflex camera kits and cookware sets, to name a few. The advancement in information technology allows more and more companies to provide customized bundles to customers. Wind and Mahajan (1997) recognize the importance of researching mass customization and suggest companies to use consumers’ input “as a response (to a conjoint analysis-type task) that provides operational guidelines for the design of products to inventory for the segment that is not willing to pay the premium required for customized products”. In addition to conjoint analysis, researchers and practitioners are using “build-your-own-bundle” or configuration approach. In a configuration study, participants are presented with a menu from which they can choose individual items to build up their desired product bundle. The process mimics the real decision process, is easy to implement, and is straight forward for participants to understand. However, as the size of the menu grows, the number of possible bundles grows geometrically. This results in computation difficulties. This dissertation investigates the application of configuration approach, and examines if it extends and complements the choice-based conjoint (CBC) approach. We first develop an aggregate model for analyzing configuration data. We show analytically that the aggregate choice model consistent with configuration data has a closed form representation which takes the form of a Multivariate Logistic (MVL) model. WE discuss the strengths and weaknesses of the configuration approach. Because configuration and conjoint data tasks have different strengths and weaknesses, taking advantages of these two choice tasks may improve the understanding of consumer preferences for bundles. A fundamental assumption in the data fusion literature is that the same decision making process is applied under different choice tasks. We examine if consumer decision making process is the same under CBC and configuration studies by comparing the estimation results from CBC and conjoint studies. We show that these two procedures may not be fully comparable. To combine the two data sources we need a data fusion model that takes into account the differences to obtain a reasonable result.

Every day, traditional statistical methodology are used world wide to study a variety of topics and provides insight regarding countless subjects. Each technique is based on a distinct set of assumptions to ensure valid results. Additionally, many statistical approaches rely on large sample behavior and may collapse or degenerate in the presence of small, spare, or correlated data. This dissertation details several advancements to detect these conditions, avoid their consequences, and analyze data in a different way to yield trustworthy results. One of the most commonly used modeling techniques for outcomes with only two possible categorical values (eg. live/die, pass/fail, better/worse, ect.) is logistic regression. While some potential complications with this approach are widely known, many investigators are unaware that their particular data does not meet the foundational assumptions, since they are not easy to verify. We have developed a routine for determining if a researcher should be concerned about potential bias in logistic regression results, so they can take steps to mitigate the bias or use a different procedure altogether to model the data. Correlated data may arise from common situations such as multi-site medical studies, research on family units, or investigations on student achievement within classrooms. In these circumstance the associations between cluster members must be included in any statistical analysis testing the hypothesis of a connection be-tween two variables in order for results to be valid. Previously investigators had to choose between using a method intended for small or sparse data while assuming independence between observations or a method that allowed for correlation between observations, while requiring large samples to be reliable. We present a new method that allows for small, clustered samples to be assessed for a relationship between a two-level predictor (eg. treatment/control) and a categorical outcome (eg. low/medium/high).

Timely identification of undesirable pavement crack conditions has been a major task in pavement management. Up to date, myriads of pavement performance models have been developed for forecasting pavement crack condition with the traditional preferred techniques being the use of regression relationships developed from laboratory and/or field statistical data. However, it becomes difficult for regression techniques to predict the crack performance accurately and robustly in the presence of a variety of tributary factors, high nonlinearity, and uncertainty. With the advancement of modeling techniques, two innovative breeds of models, Artificial Neural Networks and Markov Chains, have drawn increasing attention from researchers for modeling complex phenomena like the pavement crack performance. In this study, two distinct models, a recurrent Markov chain, and an Artificial Neural Network (ANN), were developed for modeling the performance of pavement crack condition with time. A logistic model was used to establish a dynamic relationship between transition probabilities associated with the pavement crack condition and the applicable tributary variables. The logistic model was then used conveniently to construct a recurrent Markov chain for use in predicting the crack performance of asphalt pavements in Florida. Florida pavement condition survey database were utilized to perform a case study of the proposed methodologies. For comparison purpose, a currently popular static Markov chain was also developed based on a homogeneous transition probability matrix that was derived from the crack index statistics of Florida pavement survey database. To evaluate the model performance, two comparisons were made; (1) between the recurrent Markov chain and the static Markov chain; and (2) between the recurrent Markov chain and the ANN. It is shown that the recurrent Markov chain outperforms both the static Markov chain and the ANN in terms of one-year forecasting accuracy. Therefore, with high uncertainty typically experienced in the pavement condition deterioration process, the probabilistic dynamic modeling approach as embodied in the recurrent Markov chain provides a more appropriate and applicable methodology for modeling the pavement deterioration process with respect to cracks.

[pt] Este trabalho visa analisar os modelos atuais de avaliação de risco de crédito aplicados a empresas não-financeiras e desenvolver um modelo estatístico com o emprego da ferramenta LOGIT - Regressão Logística com base nos clientes jurídicos de uma empresa do ramo industrial. Este modelo tem como objetivo principal determinar a probabilidade de um cliente ser considerado como adimplente ou inadimplente. Com esta ferramenta o analista de crédito pode definir até que ponto se torna interessante para a empresa efetuar uma venda a prazo para o cliente. / [en] This dissertation has the objective of analyzing the current models of credit risk in non financial companies and to develop a statistical model with Logistic Regression. The main purpose of this model is to determine the probability of a client (business company) being considered a good or bad risk. This model will allow the credit analyst to measure the credit risk involved with credit sales.