Up to date, it is unclear how Choice-Based Conjoint (CBC) models perform in terms of forecasting (external) real world aggregate shop data. In this contribution, we measure the performance of a Latent Class CBC model - not with an experimental holdout sample - but with aggregate real world scanning data. We find that the CBC model does not accurately predict real world market shares. In order to improve the forecasting performance, we propose a correction scheme based on external scanner data. Our analysis based on 8 brands shows that the use of the proposed correction vector improves the performance measure considerably. (author's abstract) / Series: Working Papers SFB "Adaptive Information Systems and Modelling in Economics and Management Science"

In this paper we head for a fully Bayesian analysis of the latent class model with a priori unknown number of classes. Estimation is carried out by means of Markov Chain Monte Carlo (MCMC) methods. We deal explicitely with the consequences the unidentifiability of this type of model has on MCMC estimation. Joint Bayesian estimation of all latent variables, model parameters, and parameters determining the probability law of the latent process is carried out by a new MCMC method called permutation sampling. In a first run we use the random permutation sampler to sample from the unconstrained posterior. We will demonstrate that a lot of important information, such as e.g. estimates of the subject-specific regression coefficients, is available from such an unidentified model. The MCMC output of the random permutation sampler is explored in order to find suitable identifiability constraints. In a second run we use the permutation sampler to sample from the constrained posterior by imposing identifiablity constraints. The unknown number of classes is determined by formal Bayesian model comparison through exact model likelihoods. We apply a new method of computing model likelihoods for latent class models which is based on the method of bridge sampling. The approach is applied to simulated data and to data from a metric conjoint analysis in the Austrian mineral water market. (author's abstract) / Series: Forschungsberichte / Institut für Statistik

Doctor of Philosophy / Department of Agricultural Economics / Jason S. Bergtold / The adoption of new technologies on-farm is affected by socio-economic, risk management behavior, and market factors. The adoption of cellulosic biofuel feedstock enterprises and conservation practices plays an important role in the future of Kansas agriculture. No set markets currently exist for bioenergy feedstocks and farmers may be reluctant to produce the feedstocks without contracts to mitigate uncertainty and risk. Adoption of conservation practices to improve soil productivity and health may be affected by risk considerations also. The purpose of this dissertation is to study how market mechanisms and risk influence Kansas farmers’ willingness to adopt cellulosic biofuel feedstock enterprises and conservation practices on-farm. The first essay examines farmers’ willingness to grow switchgrass under contract using a stated choice approach. Data were collected using an enumerated survey of Kansas farmers and analyzed using latent class logistic regression models. Farmers whose primary enterprise is livestock are less inclined to grow switchgrass. In addition, shorter contracts, greater harvest flexibility, crop insurance, and cost-share assistance increase the likelihood farmers will grow switchgrass. The second essay examines how farmers’ risk perceptions impact conservation practice adoption. Factor analysis of survey data was used to identify primary risk management behaviors of Kansas farmers. A multinomial logit model of conservation practice adoption incorporating these risk behaviors was developed. Estimation results indicate that different risk management factors may have no significant impact on practice adoption. Farmers may not consider certain aspects of risk significant in their adoption decision. The third essay examines the effect of different risk management behaviors on farmers’ willingness to produce alternative cellulosic bioenergy feedstocks under contract. Data were collected using a farmer survey with a set of stated choice experiments and analyzed using factor analysis and latent class logistic regression models. While farmers approach risk management differently, the risk management behaviors identified have no significant impact on farmers’ willingness to produce corn stover and switchgrass but have a negative impact on farmers’ willingness to produce sweet sorghum as a biofuel feedstock. These results may indicate that farmers are indifferent toward adopting new bioenergy cropping enterprises when traditional crop production is profitable and more certain.

This paper presents the fully Bayesian analysis of the latent class model using a new approach towards MCMC estimation in the context of mixture models. The approach starts with estimating unidentified models for various numbers of classes. Exact Bayes' factors are computed by the bridge sampling estimator to compare different models and select the number of classes. Estimation of the unidentified model is carried out using the random permutation sampler. From the unidentified model estimates for model parameters that are not class specific are derived. Then, the exploration of the MCMC output from the unconstrained model yields suitable identifiability constraints. Finally, the constrained version of the permutation sampler is used to estimate group specific parameters. Conjoint data from the Austrian mineral water market serve to illustrate the method. (author's abstract) / Series: Report Series SFB "Adaptive Information Systems and Modelling in Economics and Management Science"

In this paper we head for a fully Bayesian analysis of the latent class model with a priori unknown number of classes. Estimation is carried out by means of Markov Chain Monte Carlo (MCMC) methods. We deal explicitely with the consequences the unidentifiability of this type of model has on MCMC estimation. Joint Bayesian estimation of all latent variables, model parameters, and parameters determining the probability law of the latent process is carried out by a new MCMC method called permutation sampling. In a first run we use the random permutation sampler to sample from the unconstrained posterior. We will demonstrate that a lot of important information, such as e.g. estimates of the subject-specific regression coefficients, is available from such an unidentified model. The MCMC output of the random permutation sampler is explored in order to find suitable identifiability constraints. In a second run we use the permutation sampler to sample from the constrained posterior by imposing identifiablity constraints. The unknown number of classes is determined by formal Bayesian model comparison through exact model likelihoods. We apply a new method of computing model likelihoods for latent class models which is based on the method of bridge sampling. The approach is applied to simulated data and to data from a metric conjoint analysis in the Austrian mineral water market. (author's abstract) / Series: Working Papers SFB "Adaptive Information Systems and Modelling in Economics and Management Science"

<p>In diagnostic accuracy studies, sensitivity and specificity are the most common measures to assess the performance of diagnostic or screening tests. The estimation of these measures can be done using empirical or model-based methods. The primary objective of this thesis is to use both the empirical and the model-based (logistic regression) approach to assess the effects of covariates on the performance of the visual inspection with acetic acid (VIA) and lugol iodine (VILI) tests using the data from women screened for cervical cancer in Kinshasa, the Democratic Republic of Congo. The secondary objectives are: first, to adjust for the false negative and false positive error rates by the two tests through latent class models (LCM), and second, to evaluate the effects of covariates on the agreement between the measurements of the two tests taken by nurse and physician through Kappa statistic.</p> <p>No particular pattern could be observed in the trend of empirically estimated sensitivity and specificity of the VIA and VILI tests measured by the nurse and by the physician across age and parity categories. From the logistic regression models, both age, parity, and their respective quadratic terms have significant effects on the probability of VIA and VILI tests to detect cervical cancer. However, there is no significant effect of marital status, smoking, and hybrid capture2 (HPV DNA) on the probability of VIA and VILI tests measured by nurse to detect cervical cancer while HPV DNA does in the probability of VIA and VILI tests measured by physician to detect cervical cancer. The trend of the estimated sensitivity of VIA and VILI tests measured by the nurse is not different across age groups but the specificity does vary. The trend of both the sensitivity and specificity of VIA and VILI tests are significantly different across parity groups. The reverse is the case for the sensitivity and specificity of VIA and VILI tests measured by physician across age and parity groups. The false negative and false positive error rates in the sensitivity and specificity of VIA and VILI tests measured by nurse are higher compared to that of physician. With Kappa statistic results, there is almost perfect agreement between the ratings by the nurse and physician for the dichotomized VIA and VILI test outcomes.</p> <p>In conclusion, there is a significant effects of age, parity and the quadratic term of age on the performance of VIA and VILI tests outcomes measured by nurse. On the VIA and VILI test outcomes measured by physician, age, parity, HPV DNA and quadratic term of age have shown significant effects on the performance of VIA and VILI tests outcomes measured by physician alone.</p> / Master of Science (MSc)

Nous présentons une nouvelle méthode pour étudier les relations entre la structure génétique des populations et l'environnement. Cette méthode repose sur des modèles hiérarchiques bayésiens qui utilisent conjointement des données génétiques multi-locus et des données spatiales, environnementales et/ou culturelles. Elle permet d'estimer la structure génétique des populations, d'évaluer ses liens avec des covariables non génétiques, et de projeter la structure génétique des populations en fonction de ces covariables. Dans un premier temps, nous avons appliqué notre approche à des données de génétique humaine pour évaluer le rôle de la géographie et des langages dans la structure génétique des populations amérindiennes. Dans un deuxième temps, nous avons étudié la structure génétique des populations pour 20 espèces de plantes alpines et nous avons projeté les modifications intra spécifiques qui pourront être causées par le réchauffement climatique. / We introduce a new method to study the relationships between population genetic structure and environment. This method is based on Bayesian hierarchical models which use both multi-loci genetic data, and spatial, environmental, and/or cultural data. Our method provides the inference of population genetic structure, the evaluation of the relationships between the structure and non-genetic covariates, and the prediction of population genetic structure based on these covariates. We present two applications of our Bayesian method. First, we used human genetic data to evaluate the role of geography and languages in shaping Native American population structure. Second, we studied the population genetic structure of 20 Alpine plant species and we forecasted intra-specific changes in response to global warming. STAR

Les études diagnostiques correspondent à l’ensemble des études cliniques qui ont pour objectif l’évaluation d’un nouveau test diagnostique. Dans la démarche d’évaluation, l’étape centrale est l’évaluation de la performance du nouveau test par estimation de sa sensibilité et de sa spécificité. De manière classique, la performance du nouveau test est évaluée par comparaison à un test de référence supposé parfait, appelé un « gold standard » qui permet la connaissance du statut réel de chaque sujet vis-à-vis de la pathologie étudiée. Cependant, dans de très nombreuses situations cliniques, différentes difficultés existent : l’absence de gold standard parfait, l’impossibilité de réalisation du gold standard à tous les sujets, la dépendance des résultats des tests réalisés, la variabilité de la sensibilité et/ou de la spécificité du test en fonction de certaines conditions de réalisation, la multiple réalisation du test dans le temps ou sa multiple interprétation.Une revue méthodologique systématique a été effectuée pour faire l’état des lieux des méthodes d’inférence bayésienne disponibles dans les études diagnostiques et de leur utilisation en pratique. Le focus sur les méthodes bayésiennes a été retenu du fait de leurs avantages théoriques contrastant avec leur relative sous-utilisation dans le domaine médicale. Actuellement, de nombreuses méthodes ont été proposées pour répondre à ces différentes difficultés, avec des développements très complexes en cas de combinaison de plusieurs difficultés dans une même situation. Nous avons ainsi pu établir une cartographie des combinaisons de méthodes disponibles. Cependant leur utilisation en clinique reste encore limitée, même si elle est en augmentation ces dernières années.En pratique, nous avons été confrontés à la problématique du diagnostic de pneumopathie à Pneumocystis jirovecii (PJ) (champignon ubiquitaire opportuniste responsable de mycose profonde chez les patients immunodéprimés). Dans ce projet, nous disposions des résultats de quatre techniques de PCR (Polymerase chain reaction) différentes mais sans gold standard, avec la difficulté supplémentaire de dépendance conditionnelle entre les tests du fait du principe commun à l’origine de ces quatre tests. Deux développements ont été réalisés en parallèle pour répondre à cette problématique : d’une part, un travail sur les méthodes d’élicitation des informations a priori adaptées spécifiquement aux études diagnostiques, et d’autre part, un travail de mise en œuvre d’un modèle statistique adapté à la problématique de quatre tests dépendants en l’absence de gold standard. En l’absence de données informatives dans la littérature, l’élicitation des a priori, étape obligatoire pour l’utilisation des méthodes d’inférence bayésienne, est réalisée par l’interrogation d’experts du domaine. Notre travail a consisté en une adaptation des méthodes existantes, disponibles dans le domaine des essais cliniques, spécifiquement aux études diagnostiques pour obtenir des a priori informatifs. Cette méthode a été appliquée à notre cas des PCR diagnostiques pour PJ. L’estimation des performances diagnostiques des tests en l’absence de gold standard repose de manière efficiente sur les modèles à classes latentes. Trois modèles ont été développés pour le cas de deux tests diagnostiques : un modèle à indépendance conditionnelle, un modèle à dépendance conditionnelle à effets fixes et un modèle à dépendance conditionnelle à effets aléatoires. Nous proposons dans cette thèse une adaptation de ces trois modèles à la situation de quatre tests diagnostiques avec une formulation des paramètres permettant une interprétation clinique des covariances entre les tests dans un souci de transmission des méthodes de la théorie à la pratique. Une application et une comparaison de ces modèles ont été faites pour l’estimation des sensibilités et spécificités des quatre techniques de PCR à PJ en utilisant les a priori informatifs obtenus auprès des experts. / Diagnostic studies include all clinical studies the aim of which is the evaluation of a new diagnostic test. In the evaluation process, the main step is the evaluation of the performance of the new test i.e. its sensitivity and specificity. Usually, the performance of a new test is assessed by comparison to a test of reference which is supposed to be perfect, i.e. a "gold standard", and specifies the actual patient’s status for the disease of interest (“Diseased” or “Not-Diseased” status). However, in many clinical situations, different pitfalls exist such as (i) a gold standard is not available, (ii) the gold standard is not applicable to all patients, (iii) a conditional dependence exists between test results, (iv) the performance of a test is not constant and depends on the conditions of achievement of the test, (v) the tests are repeated in time or by several machines or read by several readers, together with multiple interpretation of the results. A systematic methodological review has been performed to inventory all Bayesian inference methods available in the field of diagnostic studies and their use in practice. The focus on Bayesian methods was based on the theoretical advantages of these methods contrasting with their relative underutilization in the medical field. Finally, several interesting methods have been proposed to address methodological issues of diagnostic studies, with very complex developments when several issues were combined in the same clinical situation. We propose to map the development methods and combinations that have already been done or not. However, their clinical use is still limited, although it has increased in recent years.In practice, we met the problem of the diagnosis of pneumonia due to Pneumocystis jirovecii (PJ). PJ is an ubiquitous opportunistic fungus leading to deep mycosis in immunocompromised patients. In this study, the results of four PCR (polymerase chain reaction) assays were available, but without any gold standard, and the supplementary difficulty of conditional dependence between tests because the four tests were based on the same principle. Two works were performed in parallel to address this issue: on one hand, an adaptation of methods to elicit prior information specifically in diagnostic studies, and on the other hand, the implementation of specific Bayesian statistical models adapted to the context of four-dependent tests in the absence of gold standard. When informative information is not available in the literature, the elicitation of priors, the mandatory first step of a Bayesian inference, is carried out by registering experts’ beliefs in the field. Our work consisted in an adaptation of existing methods, available in clinical trials, specifically for diagnostic studies to obtain informative priors. We then applied this method to our four PJ PCR assays. Estimation of the diagnostic test performance in absence of gold standard is efficiently based on latent class models (LCM). Three LCM were developed for the case of two diagnostic tests: a simple LCM assuming conditional independence between tests, a fixed effects LCM and a random effects LCM providing an adjustment for conditional dependence between tests. We extended these three models to a situation where four diagnostic tests are involved and proposed a formulation that enables an interpretation of between tests covariances in a clinical perspective in order to bind theory to practice. These models were then applied and compared in an estimation study of the sensitivities and specificities of the four PJ PCR assays, by using informative priors obtained from experts.