The aim of this work is to decide the real usage of logistic regression in the market research tasks respecting the needs of final users of research results. The main argument for the final decision is the comparison of its output to the output of an alternative classification method used in practice -- a classification tree method. The topic is divided into three parts. The first part describes the theoretical framework and approaches linked to logistic regression (chapter 2 and 3). The second part analyses the experience with the usage of logistic regression in Czech market research companies (chapter 4) and the topic is closed by applying the method on real data and comparing the output to the classification tree output (chapter 5 and 6).

[EN] The Statistical Process Control (SPC) is a method used to control the quality characteristics of a product during the production process, determine whether the manufacturing process is or not stable and improve its capacity through the reduction of variability. One of the main tools used in the SPC is the control chart. Often the quality of a product is measured through various quality characteristics generally correlated. Multivariate Control charts are a response to the need for quality control in such situations. If the quality characteristics are qualitative, sometimes it happens that the product quality is defined by linguistic variables and product units are also classified by linguistic forms into several categories, depending on the degree of fulfillment of expectations, creating a situation of fuzzy classifications. The control charts proposed in the literature to deal with such situations are mostly based on simulation and using approximation techniques which hinder the practical application thereof. This thesis addresses this issue proposing a multivariate control chart for quality characteristics of multi-type attributes correlated based on the T2 control chart of Hotelling, using a fuzzy approach. The results of the proposed control charts before are improved by establishing a more formal way of measuring and evaluating quality in these diffuse situations. A method is also proposed to assess the performance of control chart proposed, by deter mining the average run length (ARL), in both in-control state and the out-of-control state. For this, algorithms which use Monte Carlo simulation are developed and implemented in R. Additionally, the sensitivity of the control chart faced with the choice of the membership functions of linguistic variables is analyzed. / [ES] El Control Estadístico de Procesos (CEP) es un método que se utiliza para controlar las características de calidad de un producto durante el proceso de producción, determinar si los procesos de manufactura son o no estables y mejorar su capacidad a través de la reducción de la variabilidad. Una de las principales herramientas utilizadas en el Control Estadístico de Procesos es el gráfico de control. Con frecuencia, la calidad de un producto se mide a través de varias características de calidad, generalmente correlacionadas. Los gráficos de control multivariantes son una respuesta a la necesidad de controlar la calidad en tales situaciones. Si las características de calidad son de carácter cualitativo, ocurre en ocasiones que la calidad del producto se define mediante variables lingüísticas y las unidades de producto se clasifican también de for ma lingüística en varias categorías, dependiendo del grado de cumplimiento de las expectativas, creando una situación de clasificaciones difusas. Los gráficos propuestos en la literatura para tratar con tales situaciones están, en su mayoría, basados en simulación y el uso de técnicas de aproximación que dificultan la aplicación práctica de los mismos. Esta tesis trata esta cuestión proponiendo un Gráfico de Control multivariante para características de calidad de tipo multi-atributos correlacionados basado en el gráfico T2 de Hotelling, utilizando un enfoque difuso. Se mejora los resultados de los gráficos de control propuestos anterior mente estableciendo una manera más formal de medición y evaluación de la calidad en estas situaciones difusas. Se propone además un procedimiento para evaluar el rendimiento del gráfico de control propuesto mediante la determinación de la longitud de racha promedio (ARL), tanto para un estado bajo-control como para el estado fuera-de-control. Para ello se desarrollaron algoritmos que utilizan simulación de Monte Carlo y han sido implementados en R. Adicionalmente, se analiza la sensibilidad del gráfico de control frente a la elección de las funciones de pertenencia de las variables lingüísticas. / [CAT] El Control Estadístic de Processos (CEP) és un mètode que s'utilitza per controlar les característiques de qualitat d'un producte durant el procés de producció, deter minar si els processos de manufactura són o no estables i millorar la seva capacitat a través de la reducció de la variabilitat. Una de les principals eines utilitzades en el Control Estadístic de Processos és el gràfic de control. Sovint, la qualitat d'un producte es mesura a través de diverses característiques de qualitat, generalment correlacionades. Els gràfics de control multivariants són una resposta a la necessitat de controlar la qualitat en aquestes situacions. Si les característiques de qualitat són de caràcter qualitatiu, de vegades passa que la qualitat del producte es defineix mitjançant variables lingüístiques i les unitats de producte es classifiquen també de for ma lingüística en diverses categories, depenent del grau de compliment de les expectatives, creant una situació de classificacions difuses. Els gràfics proposats en la literatura per abordar aquestes situacions són, majoritàriament, basats en simulació i l'ús de tècniques d'aproximació que en dificulten l'aplicació pràctica. Aquesta tesi tracta de resoldre aquesta qüestió amb la proposta d'un Gràfic de Control multivariant per característiques de qualitat de tipus multi-atributs correlacionats basat en el gràfic T2 de Hotelling, mijançant un enfocament difús. S'hi milloren els resultats de les gràfics de control proposats anterior ment per mitjà d'un mètode més for mal de mesurament i avaluació de la qualitat en aquestes situacions difuses. S'hi proposa a més un procediment per avaluar el rendiment del gràfic de control proposat mitjançant la deter minació de la longitud de ràfega mitjana (ARL), tant per a un estat en-control com per a l'estat fora-de-control. Amb aquesta finalitat es van desenvolupar algoritmes que utilitzen simulació de Monte Carlo i han estat implementats en R. Addicionalment, s'hi analitza la sensibilitat del gràfic de control davant l'elecció de les funcions de pertinença de les variables lingüístiques. / Pastuizaca Fernández, MN. (2016). Diseño y mejora de gráficos de control multivariantes para atributos. Un enfoque basado en teoría difusa [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/65073 / TESIS

Le premier chapitre, écrit conjointement avec Alfred Galichon, Keith O'Hara et Matthew Shum, montre l'équivalence entre les modèles de choix discrets et les modèles d'appariements. Cette équivalence permet l'estimation efficace, par des algorithmes d'appariement, de modèles qui étaient jusqu'à présent réputés comme difficile à estimer dans la littérature. Le deuxième chapitre, écrit conjointement avec Mathilde Poulhès, s'appuie sur les résultats du premier pour estimer le consentement marginal à payer des agents pour différentes caractéristiques du logement et du quartier à Paris. Il introduit une nouvelle procédure d'estimation basée sur le modèle de pures caractéristiques. Grâce à un riche jeu de données sur les achats de logements à Paris, nous montrons que le revenu moyen du voisinage et le niveau de criminalité sont de puissants déterminants du choix du quartier pour tous les types d'acheteurs, que l'accessibilité à l'emploi est également un facteur déterminant pour les ménages comptant plus d'une personne, et que la qualité de l'école du secteur joue un rôle primordial pour les ménages avec enfants. Le troisième chapitre, écrit conjointement avec Guillaume Chapelle, Alain Trannoy et Etienne Wasmer, montre que la croissance récente du ratio patrimoine sur revenu est due uniquement à l'augmentation du prix des logements, et plus précisément à l'augmentation du prix d'un facteur fixe de production: la terre. Nous montrons ensuite qu'un système de taxation du patrimoine doit taxer le facteur fixe qu'est la terre à des fins de redistribution et non le capital productif pour ne pas décourager l'investissement. / The first two of the three chapters of this thesis examine the identification and the estimation of discrete choice models. The first chapter proves the equivalence between matching models and discrete choice models, and draws the consequences in terms of identification and estimation. The second chapter builds on the results of the first, and uses matching algorithms to estimate the marginal willingness to pay of households for various housing and neighborhood characteristics in Paris (such as school performance, crime level, distance to employment areas). The third chapter deals with another topic: it first shows that the recent rise in the capital-income ratio highlighted by Thomas Piketty in his book is due to the rise in housing prices, and it then explores the consequences in terms of wealth distribution.

E-Commerce research shows that existing studies on online consumer choice behavior has focused on comparative studies of channel or store choice (online or offline), or online store choice (different e-tailers). Relatively less effort has been devoted to consumers’ online brand choice behavior within a single e-tailer. The goal of this research is to model online brand choice, including generating loyalty variables, setting up base model, and exploring the effects of Internet-specific attributes, i.e., order delivery, webpage display and order confirmation, on online brand choice at the SKU level. Specifically, this research adopts the Multinomial Logit Model (MNL) as the estimation methods. To minimize the model bias, the refined smoothing constants for loyalty variables (brand loyalty, size loyalty, and SKU loyalty) are generated using the Nonlinear Estimation Algorithm (NEA). The findings suggest that SKU loyalty is a better predictor of online brand choice than brand loyalty and size loyalty. While webpage display has little effect on the brand choice, order delivery has positive effect on the choice. Online order confirmation turns out to be helpful in choice estimation. Moreover, online consumers are not sensitive to net price of the alternatives, but quite sensitive to price promotion. These results have meaningful implications for marketing promotions in the online environment and suggestions for future research.

Thesis (M.Sc. (Statistics)) -- University of Limpopo, 2015 / Introduction: The high number of variables from the 2009 General Household Survey is prohibitive to do holistic analysis of data due to high correlations that exist among many variables, making it virtually impractical to apply traditional methods such as multinomial logistic regression. The purpose of this study to identify observed variables that can be explained by a few unobservable quantities called factors, using factor analysis. Methods: Factor analysis is used to describe covariance relationships among 162 variables of interest in the 2009 General Household Survey (GHS) and 2009 Quarterly Labour Force Survey of South Africa (QLFS). Data for the respondents aged 15 years and above was analysed by first applying factor analysis to the 162 variables to produce factor scores and develop models for five core areas: education, health, housing, labour force and social development. Multinomial logistic regression was then used to model educational levels and service satisfaction using identified factor sores. Results: The variability among the 162 variables of interest was described by only 29 factors identified using factor analysis, even though these factors are not measured directly. Multinomial logistic regression (MLR) analysis showed negative and significant impact of education factors (fees too high, violence and absence of parental care) on levels of educational attainment. “Historically advantaged” factor is the only factor significant and positively affects educational levels. Housing and social development factors were regressed against service satisfaction. Housing factors such as the home owners, age of a house and male household heads were found to be significant. Social development factors such as “no problem with health”, sufficient water, high income, household size and telephone access were found to be significant. Labour force factors such as employment, industrial business and occupation, employment history and long-term unemployment have positive and significant impact on levels of education. Conclusion: It can be concluded that factor analysis as a data reduction technique has managed to describe the variability among the 162 variables in terms of just 29 unobservable variables. Using MLR in subsequent analysis, this study has managed to identify factors positively or negatively associated with educational levels and service satisfaction. The study suggests that educational, housing, social development and labour force facilities should be improved and education should be used to improve life circumstances. Keywords: factor analysis, factors, multinomial logistic regression, logits, educational levels of attainment, service satisfaction, quality of service delivery. / DST-NRF, Centre of Excellence in Mathematical and Statistical Sciences (MaSS)

The transportation sector is considered one of the major contributors to greenhouse gas (GHG) emissions in metropolitan areas, and any efforts to reduce these emissions requires strategic management of multiple transportation modes. This paper presents a method to identify opportunities to reduce GHG emissions by expanding commuter bus services and incentives to shift commuters from private cars to transit. The approach uses a nested multinomial logit model for mode choice in a region that includes driving alone, carpooling, walking, cycling, and using four possible transit modes (ferry, commuter rail, rapid transit and bus) by walk access or driving access. A model of existing conditions was calibrated with data from the Boston metropolitan area. Using an emission factor model based on average speeds from the California Air Resources Board (CARB), the net effect of new commuter bus service on GHG emissions from transportation was estimated. Potential GHG reductions are weighed against the capital and operating costs of new transit services to quantify the cost-effectiveness of a new commuter bus service for isolated origin-destination pairs. This modeling framework is used to optimize fares and bus frequency in order to identify the corridors with the most cost-effective potential for GHG reduction. Results are presented for the Boston region, demonstrating the feasibility of implementation and the potential magnitude of benefits for cost-effectively reducing GHG emissions associated with transportation. The method is general and can be applied in other cities around the world.

The main goal of this thesis is to analyse the fluctuation of the employees in a well established Czech manufacturing company. Due to the GDPR regulations, the underlying company is kept anonymised in this thesis. The data were transformed into longitudinal data and the GEE methodology was used for the analysis of the fluctuation. In the first chapter, an introduction to the problem and a short description of the data is provided. The second chapter contains some theoretical description of the GEE methodology and the QIC information criterion. In the third chapter, multiple models for a binary and multinomial response are fitted to the data and their results are described in detail. This allows us to describe the influence of various factors to the fluctuation of the employees in the underlying company. 1

Social media has become an integral part of the Internet. There have been users across the world sharing content like images, texts, videos, and so on. There is a huge amount of data being generated and it has become a challenge to the social media platforms to group the content for further usage like recommending a video. Especially, grouping videos based on similarity requires extracting features. This thesis investigates potential approaches to extract features that can help in determining the similarity between videos. Features of given videos are extracted using Object Detection and Action Recognition. Bag-of-features representation is used to build the vocabulary of all the features and transform data that can be useful in clustering videos. Probabilistic model-based clustering, Multinomial Mixture model is used to determine the underlying clusters within the data by maximizing the expected log-likelihood and estimating the parameters of data as well as probabilities of clusters. Analysis of clusters is done to understand the genre based on dominant actions and objects. Bayesian Information Criterion(BIC) and Akaike Information Criterion(AIC) are used to determine the optimal number of clusters within the given videos. AIC/BIC scores achieved minimum scores at 32 clusters which are chosen to be the optimal number of clusters. The data is labeled with the genres and Logistic regression is performed to check the cluster performance on test data and has achieved 96% accuracy

Arbuscular mycorrhizal (AM) fungi are key soil symbiotic microorganisms, intensively studied for their roles in improving plant fitness and their ubiquity in terrestrial ecosystems. Research on AM fungi is difficult because their obligate biotrophic nature makes it impossible to culture them in the absence of a host. Over the last three decades, Ri T-DNA transformed roots have been the gold standard to study AM fungi under in vitro conditions. However, only two host plant species (Daucus carota and Cichorium intybus) have been routinely used to in vitro propagate less than 5% of the known AM fungal species. There is much evidence that host identity can significantly affect AM symbioses, therefore, we investigated any potential host-specific effects of two novel Ri T-DNA transformed root species, Medicago truncatula and Nicotiana benthamiana, by associating them with seven AM fungal species selected based on their contrasting behaviors when grown with Ri T-DNA transformed D. carota roots. To evaluate the performance of new Ri T-DNA transformed roots to host and propagate AM fungal species, a factorial set-up was used to generate nine unique pairs of hosts (M. truncatula, N. benthamiana, D. carota) and AM fungi (Rhizophagus irregularis, R. clarus, Glomus sp.). Using statistical modeling, all pairs of hosts and AM fungi were compared by their symbiosis development (SD) and sporulation patterns in the hyphal compartments (HCs) of two-compartment Petri dishes. Our results show that 1) most of the variation between host and AM fungus pairs relating to SD or HC sporulation was explained by an interaction between host and AM fungal identity, i.e., host identity alone was not sufficient to explain AM fungal behaviour, 2) AM symbioses involving different combinations of symbiont identities trigger heterogenous fungal behaviours. This work provides a robust framework to develop and evaluate new Ri T-DNA roots for the in vitro propagation of AM fungi, an important asset for germplasm collections and biodiversity preservation.

京都大学 / 0048 / 新制・課程博士 / 博士(経済学) / 甲第19459号 / 経博第528号 / 新制||経||276(附属図書館) / 32495 / 京都大学大学院経済学研究科経済学専攻 / (主査)教授 西山 慶彦, 准教授 奥井 亮, 准教授 高野 久紀 / 学位規則第4条第1項該当 / Doctor of Economics / Kyoto University / DGAM