Search

The study is aimed to construction of top-soil salinity maps for three coastal districts of Ben Tre province, which is one of the Mekong Delta areas that have been seriously affected by saline intrusion in recent years. A total of 124 top-soil samples were collected and analyzed during the dry season in 2021, of which 59 and 65 samples were taken in January and March respectively. Multivariable linear regression model was then built based on measured salinity, by Electrical Conductivity parameter, and spectral bands extracted from Landsat 8 satellite images for the respective days. The model was built based on the Bayesian statistical method, processed by the R statistical program (version 4.0.3), which big difference between Bayesian and classical statistical methods is that this method does not need to evaluate the distribution of the inputs, therefore helping to optimize the quality and quantity of the collected dataset. The results of the model showed that the model efficiency reaches 67%, and the RMSE and IOA validation parameters meet the requirements: 1.31 and 0.63 respectively. The model also showed that the infrared and panchromatic spectral bands are the two spectral bands which are capable of predicting the salinity of the topsoil in this area. Therefore, the results of this study contribute to a potential method in building an effective and low-cost model to assess surface soil salinity in the study area.

The cingulate cortex is known to be a complex structure, involved in several cognitive and emotional functions, as well as being altered by a variety of brain disorders. This heterogeneity is reflected in the multiple parceling models proposed in the literature. At the present, sub-regions of the cingulate cortex had been identified taking into account functional and structural connectivity, as well as cytological and electrochemical properties. In the present work, we propose an innovative node-wise parceling approach based on meta-analytic Bayesian co-alteration. To this aim, 193 case–control voxel-based morphometry experiments were analyzed, and the Patel’s κ index was used to assess probability of morphometric co-alteration between nodes placed in the cingulate cortex and in the rest of the brain. Hierarchical clustering was then applied to identify nodes in the cingulate cortex exhibiting a similar pattern of whole-brain co-alteration. The obtained dendrogram highlighted a robust fronto-parietal cluster compatible with the default mode network, and being supported by the interplay between the retrosplenial cortex and the anterior and posterior cingulate cortex, rarely described in the literature. This ensemble was further confirmed by the analysis of functional patterns. Leveraging on co-alteration to investigate cortical organization could, therefore, allow to combine multimodal information, resolving conflicting results sometimes coming from the separate use of singular modalities. Crucially, this provides a valuable way to understand the pathological brain using data driven, whole-brain informed and context-specific evidence in a way not yet explored in the field.

International audience; We study functional inequalities (Poincar\'e, Cheeger, log-Sobolev) for probability measures obtained as perturbations. Several explicit results for general measures as well as log-concave distributions are given.The initial goal of this work was to obtain explicit bounds on the constants in view of statistical applications for instance. These results are then applied to the Langevin Monte-Carlo method used in statistics in order to compute Bayesian estimators.

BACKGROUND: Delirium is highly prevalent in the intensive care unit (ICU) and is associated with high morbidity and mortality. The antipsychotic haloperidol is the most frequently used agent to treat delirium although this is not supported by solid evidence. The agents intervening against delirium in the intensive care unit (AID-ICU) trial investigates the effects of haloperidol versus placebo for the treatment of delirium in adult ICU patients.METHODS: This protocol describes the secondary, pre-planned Bayesian analyses of the primary and secondary outcomes up to day 90 of the AID-ICU trial. We will use Bayesian linear regression models for all count outcomes and Bayesian logistic regression models for all dichotomous outcomes. We will adjust for stratification variables (site and delirium subtype) and use weakly informative priors supplemented with sensitivity analyses using sceptical priors. We will present results as absolute differences (mean differences and risk differences) and relative differences (ratios of means and relative risks). Posteriors will be summarised using median values as point estimates and percentile-based 95% credibility intervals. Probabilities of any benefit/harm, clinically important benefit/harm and clinically unimportant differences will be presented for all outcomes.DISCUSSION: The results of this secondary, pre-planned Bayesian analysis will complement the primary frequentist analysis of the AID-ICU trial and facilitate a nuanced and probabilistic interpretation of the trial results.

The production of platform molecules from the valorization of lignocellulosic biomass is increasing. Among these plateform molecules, γ-valerolactone (GVL) is a promising one and could be used for different industrial applications. This molecule is synthesized from levulinic acid (LA) or alkyl levulinates (AL) through a tandem hydrogenation/cyclization (lactonization) cascade. A lot of investigations have been carried out to develop the best catalyst for the hydrogenation step by using solely LA or AL. However, one should keep in mind that in the AL production via fructose alcoholysis, there is also LA production, and both are present in the product mixture during the further conversion. To the best of our knowledge, no article exists describing the hydrogenation of LA and AL simultaneously in one-pot. Also, the literature reporting the use of solid catalyst for the second cyclization step is rare. To fill this gap, the hydrogenation of levulinic acid and butyl levulinate was studied over Ru/C and Amberlite IR-120. Several kinetic models were evaluated via Bayesian inference and K-fold approach. The kinetic assessment showed that a non-competitive Langmuir-Hinshelwood with no dissociation of hydrogen where LA, BL and H2 are adsorbed on different sites (NCLH1.2) and non-competitive Langmuir-Hinshelwood with dissociation of hydrogen where LA, BL (butyl levulinate) and H2 are adsorbed on different sites (NCLH2.2) are the best model to describe this system. The presence of LA and Amberlite IR-120 allows to increase the kinetics of cyclization steps, and in fine to accelerate the production of GVL.

This thesis takes place in extreme value statistics and agricultural insurance frameworks. For the first line of research, the extreme quantile of a response variable Y can often be linked to a vector of covariates X of dimension p. When p is large compared to the sample size n, the conditional distribution of Y given X becomes difficult to estimate, especially when dealing with extreme values. The first contribution of this thesis is to propose a new approach, called Extreme-PLS, for dimension reduction in conditional extreme values settings. This approach consists in reducing the dimension of X by maximizing the covariance between a linear combination of coordinates X and Y given large values of Y. We establish the asymptotic normality of the Extreme-PLS estimator under a single-index model. The second contribution provides a Bayesian extension to the Extreme-PLS method to address data scarcity problems in distribution tails. This approach allows to identify the direction of dimension reduction by introducing a prior information on it. It provides a Bayesian framework for computing the posterior distribution of the direction, where the likelihood function is obtained from a von Mises-Fisher distribution adapted to hyperballs. Three prior distributions are considered: conjugate, hierarchical and sparse priors. Finally, the performance of both approaches is evaluated on simulated data, and an application on French farm income data is provided as an illustration.Regarding the second line of research, climate disruption and market deregulation have increased and impacted agricultural production and income. Farmers' incomes are faced with two main types of risk related to price and yield volatility. Protection against these risks fall within a good risk management and thus farmers' insurance coverage. The third contribution of this thesis concerns the study and modelling of the dependence structure between crop yield and price risks using copulas. We also use conditional copulas to take into account the effect of other covariates such as crop insurance purchase, claims and weather factors. The last contribution focuses on considering the natural hedge mechanism, i.e. the negative dependence between yields and prices, in a revenue insurance scheme. We analyse its effect on the value of the actuarially fair premium on an example of revenue insurance contract pricing. The results show that a natural hedge is likely to reduce insurance premiums in France.All studies focus on French farm income data in the cereal (maize and wheat) and wine sectors., Cette thèse s'inscrit dans le cadre de la statistique des valeurs extrêmes et de l'assurance agricole. Pour le premier axe de recherche, le quantile extrême d'une variable réponse Y peut souvent être lié à un vecteur de covariables X de dimension p. Lorsque p est grand comparé à la taille de l'échantillon n, la distribution conditionnelle de Y étant donné X devient difficile à estimer, surtout lorsqu'on a affaire à des valeurs extrêmes. La première contribution de cette thèse est de proposer une nouvelle approche, appelée Extreme-PLS, pour la réduction de la dimension dans le cadre des valeurs extrêmes conditionnelles. Cette approche consiste à réduire la dimension de X en maximisant la covariance entre une combinaison linéaire des composants de X et de Y étant donné de grandes valeurs de Y. Nous établissons la normalité asymptotique de l'estimateur Extreme-PLS sous un modèle à indice unique. La deuxième contribution est une extension bayésienne de la méthode Extreme-PLS pour traiter les problèmes de rareté des données dans les queues de distribution. Cette approche permet d'identifier la direction de la réduction de la dimension en introduisant des informations a priori sur celle-ci. Elle fournit un cadre bayésien pour calculer la distribution postérieure de la direction, où la fonction de vraisemblance est obtenue à partir d'une distribution de von Mises-Fisher adaptée aux hyper boules. Trois distributions postérieures sont considérées : loi conjuguée, hiérarchique et sparse.Enfin, la performance des deux approches est évaluée sur des données simulées, et une application sur des données de revenus agricoles françaises est fournie à titre d'illustration.En ce qui concerne le deuxième axe de recherche, les dérèglements climatiques et la dérégulation des marchés ont augmenté et impacté la production et les revenus agricoles. Les revenus des agriculteurs sont confrontés à deux principaux types de risques liés à la volatilité des prix et des rendements. La protection contre ces risques relève d'une bonne gestion des risques et donc de la couverture d'assurance des agriculteurs. La troisième contribution de cette thèse concerne l'étude et la modélisation de la structure de dépendance entre le rendement des cultures et les risques de prix par les copules. Nous utilisons également les copules conditionnelles pour prendre en compte l'effet d'autres covariables telles que l'achat d'assurance récolte, les sinistres et les facteurs météorologiques. La dernière contribution porte sur la prise en compte du mécanisme de couverture naturelle, c-à-d la dépendance négative entre les rendements et les prix, dans un régime d'assurance des revenus. Nous analysons son effet sur la valeur de la prime actuariellement juste sur un exemple de tarification de contrat d'assurance revenu. Les résultats montrent qu'une couverture naturelle est susceptible de réduire les primes d'assurance en France.L'ensemble des études se concentrent sur les données du revenu agricole français dans les secteurs céréaliers (maïs et blé) et viticoles.

The synthesis of γ-valerolactone (GVL), a platform molecule that can be produced from lignocellulosic biomass, was performed in this work by hydrogenation of an alkyl levulinate over Ru/C. Kinetic models reported in the literature are typically not compared with rival alternatives, even if a discrimination study is needed to find the optimum operating conditions. Different surface reaction kinetic models were thus considered in this work, specifically addressing hydrogenation of butyl levulinate to GVL, where the latter was used as a solvent to minimize potential solvent interference with the reaction, including its evaporation. The Bayesian approach was applied to evaluate the probability of each model. It was found that non-competitive Langmuir-Hinshelwood with no dissociation of the hydrogen model has the highest posterior probability.

This master thesis addresses the issue of protein cluster quantification in superresolution microscopy. Methods in super-resolution microscopy, typically referred to as single molecule localization microscopy, have greatly improved the optical resolution by sequentially exciting only the part of fluorescent molecules and image their signals through time. The signal is processed and stored as a localization in space. The improved resolution has further demonstrated molecular clustering as a relevant biological feature for cellular function. However, each cluster can be composed of several molecules and each of them is subjected to stochasticity of molecule labeling and complex photophysics of the fluorescent probes. This leads to a broad distribution of number of localizations for each cluster size, which impacts exact quantification of cluster stoichiometry. The aim of this master thesis was to investigate the performance of nested sampling compared to the previously developed method in Zanacchi et al. (2017). In the said article, the authors developed a method based on numerical approximation which estimated the total protein count as the mixture model of several oligomeric states. The proportions of each oligomeric state were estimated based on the number of localizations in each cluster. In this master thesis, we implemented the nested sampling based on the mixture model described above and compared both approaches on simulated and real data. The goal of both approaches was firstly to estimate the number of different oligomeric states in the model and secondly their corresponding proportions. The methods were evaluated in a simulation study and on the data generated from STORM imaging. The simulation study showed better performance of nested sampling especially in smaller samples, while in larger samples the methods performed similarly. In the STORM image analysis, where all the samples were large (n > 1000), all fitted distributions were almost identical. Magistrsko delo obravnava težavo kvantifikacije proteinov v super-resolucijski mikroskopiji. Metode super-resolucijske mikroskopije so bistveno izboljšale ločljivost na način, da naenkrat vzbudijo le del fluorescenčnih molekul in jih slikajo. Ta proces je izveden večkrat zaporedoma, končna slika pa je sestavljena iz vseh zaporednih slik. Vsak signal je obdelan in shranjen kot lokacija molekule v prostoru. Kljub izboljšanju ločljivosti je meja difrakcije teh metod med 20 in 50 nm, zaradi česar so molekule, ki so si bližje od te razdalje, na končni rekonstruirani sliki vidne kot gruča. Gručenje proteinov je biološka lastnost, ki nas zanima. Ker je posamezna gruča na končni sliki lahko sestavljena iz ene ali več ločenih molekul in vsaka od molekul lahko odda več signalov, kvantifikacija s preprostim štetjem posameznih gruč ni mogoča in problem ni lahko rešljiv. Namen magistrskega dela je bil raziskati uspešnost “nested sampling” metode v primerjavi s predhodno razvito metodo v Zanacchi et al. (2017). V omenjenem članku so avtorji razvili metodo, ki je skupno število proteinov ocenila kot mešani model več oligomernih stanj. Delež vsakega oligomernega stanja je bil ocenjen na podlagi števila lokalizacij v vsaki gruči. Rezultati v Zanacchi et al. (2017) so bili pridobljeni z numerično optimizacjo. V tej magistrski nalogi smo izvedli “neseted sampling” na podlagi zgoraj opisanega mešanega modela in primerjali oba pristopa na simuliranih in realnih podatkih. Cilj obeh pristopov je bil najprej oceniti število različnih oligomernih stanj in nato oceniti še njihov delež v mešanem modelu. Metode so bile ovrednotene s simulacijsko študijo in na podatkih, pridobljenih s STORM slikanjem. Simulacijska študija je pokazala boljše delovanje “neseted sampling”, zlasti pri manjših vzorcih, medtem ko pri večjih vzorcih večjih razlik med metodami ni bilo. V analizi STORM podatkov, kjer so bili vsi vzorci veliki (n > 1000), so si bile ocenjene porazdelitve iz obeh metod zelo podobne.

International audience; The use of hydrogen peroxide as an oxidizing agent becomes increasingly important in chemistry. The example of vegetable oil epoxidation is an excellent illustration of the potential of such an agent. This reaction is traditionally performed by Prileschajew oxidation, i.e., by the in situ production of percarboxylic acids. Drawbacks of this approach are side reactions of ring-opening and thermal runaway reactions due to percarboxylic acid instability. One way to overcome this issue is the direct epoxidation by hydrogen peroxide by using γ-alumina. However, the reaction mechanism is not elucidated: does hydrogen peroxide decompose with alumina or oxidize the hydroxyl groups at the surface? The kinetics of hydrogen peroxide consumption with alumina in homogeneous liquid and heterogeneous liquid-liquid systems was investigated to reply to this question. Bayesian inference was used to determine the most probable models. The results obtained led us to conclude that the oxidation mechanism is the most credible for the heterogeneous liquid-liquid system.

Liver transplantation for primary sclerosing cholangitis (PSC) can be complicated by recurrence of PSC (rPSC). This may compromise graft survival but the effect on patient survival is less clear. We investigated the effect of post-transplant rPSC on graft and patient survival in a large European cohort. Registry data from the European Liver Transplant Registry regarding all first transplants for PSC between 1980 and 2015 were supplemented with detailed data on rPSC from 48 out of 138 contributing transplant centres, involving 1,549 patients. Bayesian proportional hazards models were used to investigate the impact of rPSC and other covariates on patient and graft survival. Recurrence of PSC was diagnosed in 259 patients (16.7%) after a median follow-up of 5.0 years (quantile 2.5%-97.5%: 0.4-18.5), with a significant negative impact on both graft (HR 6.7; 95% CI 4.9-9.1) and patient survival (HR 2.3; 95% CI 1.5-3.3). Patients with rPSC underwent significantly more re-transplants than those without rPSC (OR 3.6, 95% CI 2.7-4.8). PSC recurrence has a negative impact on both graft and patient survival, independent of transplant-related covariates. Recurrence of PSC leads to higher number of re-transplantations and a 33% decrease in 10-year graft survival. ispartof: TRANSPLANT INTERNATIONAL vol:34 issue:8 pages:1455-1467 ispartof: location:Switzerland status: published

Out of Shelf availability is a major source of concern in retail sector. Many studies have approached these challenging issue from different perspectives analysing coordination between retailers and manufacturers, inventory management, estimation of consumers demand, among others. On the other hand, it is quite difficult to detect this kind of situations because it implies an important investment in work-force and/or technological devices. The purpose of this thesis is to address the problematic of Out of Shelf products in the supermarket business by differentiating regular and irregular patterns of transactions. Hidden Markov Model and Bayesian Statistic are applied in order to identify periods of time where products are not available on shelf to be purchased by shoppers, recognizing Out of Shelf states using transactional data.

We present deep $g$- and $r$-band Magellan/Megacam photometry of two dwarf galaxy candidates discovered in the Dark Energy Survey (DES), Grus I and Indus II (DES J2038-4609). For the case of Grus I, we resolved the main sequence turn-off (MSTO) and $\sim 2$ mags below it. The MSTO can be seen at $g_0\sim 24$ with a photometric uncertainty of $0.03$ mag. We show Grus I to be consistent with an old, metal-poor ($\sim 13.3$ Gyr, [Fe/H]$\sim-1.9$) dwarf galaxy. We derive updated distance and structural parameters for Grus I using this deep, uniform, wide-field data set. We find an azimuthally averaged half-light radius more than two times larger ($\sim 151^{+21}_{-31}$ pc; $\sim 4.^{\prime} 16^{+0.54}_{-0.74}$) and an absolute $V$-band magnitude $\sim-4.1$ that is $\sim 1$ magnitude brighter than previous studies. We obtain updated distance, ellipticity, and centroid parameters which are in agreement with other studies within uncertainties. Although our photometry of Indus II is $\sim 2-3$ magnitudes deeper than the DES Y1 Public release, we find no coherent stellar population at its reported location. The original detection was located in an incomplete region of sky in the DES Y2Q1 data set and was flagged due to potential blue horizontal branch member stars. The best fit isochrone parameters are physically inconsistent with both dwarf galaxies and globular clusters. We conclude that Indus II is likely a false-positive, flagged due to a chance alignment of stars along the line of sight., 16 pages, 8 figures; submitted to ApJ

International audience; The use of lignocellulosic biomass in the chemical industry can significantly contribute to respect the various international agreements on climate change. One of the most promising platform molecules issued from the lignocellulosic biomass hydrolysis is γ-valerolactone (GVL). GVL can be upgraded to valuable chemicals and produced by the hydrogenation of alkyl levulinates. Although these reactions are widely studied, seldom research focused on the solvent effect. To fill this gap, the effect of three different reaction mixtures with an excess of butyl levulinate (BL), of butanol and GVL was studied on the kinetics of BL hydrogenation to GVL over Ru/C. PC-SAFT (Perturbed-Chain Statistical Associating Fluid Theory) shows that the solubility of hydrogen is not constant during the reaction progress, and it was taken into account. To allow a fair comparison, kinetic models were developed using Bayesian statistics for each reaction mixture. The best performances were obtained when the reaction mixture has an excess of GVL.

Single case studies continue to play an important role in neuropsychological research. However, the range of statistical tools specifically designed for single cases is still limited. The current gold standard is the Crawford's t-test, but it is crucial to note that this is limited to simple designs and it is not possible to make inferences relevant to support for the null hypothesis with it. The Bayesian Multilevel Single Case models (BMSC) provide a novel tool that grants the flexibility of linear mixed model designs. BMSC is also able to support both null and alternative hypotheses in complex experimental designs using the Bayesian framework. We compared the BMSC and Crawford's t-test in a simulation study involving a case of no-dissociation and a case of simple dissociation between a single case patient and a series of control groups of different sizes (N = 5, 15, or 30). We then showed how BMSC is useful in complex designs by means of an example using real data. The BMSC proved to be more reliable than the Crawford's test, in terms of first-type errors and more precise estimating the parameters. Notably, the BMSC model provides a comprehensive vision of the whole experimental design, interpolating a single model. It follows the recent trend which involves a shift in attention from p-values to other inferential indices and estimates.

Diabetes mellitus is an alarming problem worldwide with various negative impacts inherent to the multidimensional complexity of its nature. In particular, although it is preventable, type 2 diabetes mellitus afflicts 90% of the total diabetic population. The speed with which this preventable disease is deteriorating society, overloading health systems, draining health budgets and finally slowing economic growth, has led to its being considered one of the main health emergencies of the 21st century, and to it becoming an area of significant research attention. In Mexico, efforts to address the alarming calls launched worldwide to prevent type 2 diabetes mellitus are still underway. In particular, national studies have not yet integrated multidisciplinary techniques of data science into their analysis methodology, leaving an opportunity gap open for the process of knowledge extraction. This work addresses this gap by presenting a multidisciplinary study that integrates data science tools for the national study of type 2 diabetes mellitus. The study reveals implicit, non-trivial, unknown and potentially useful patterns of information in the national study of type 2 diabetes mellitus, its main associated health variables and its socio-demographic profile. The results obtained present quality information that could be used to manage and support effective strategies to combat the prevalence of type 2 diabetes mellitus in Mexico. Doctor of Philosophy in Engineering Science

I modelli mistura bayesiani sono onnipresenti in statistica per la loro semplicità e flessibilità e possono essere facilmente impiegati in un'ampia varietà di contesti. In questa tesi, miriamo a fornire alcuni contributi agli attuali metodi bayesiani di analisi dei dati, spesso motivati ​​da domande di ricerca provenienti da applicazioni biologiche. In particolare, ci concentriamo sullo sviluppo di nuovi modelli mistura bayesiani, tipicamente in un ambiente non parametrico, per migliorare ed estendere aree di ricerca che coinvolgono dati caratterizzati da grande dimensioni: la modellazione di dati nested, test di ipotesi simultaneo e la riduzione della dimensionalità. \\ Pertanto, il nostro obiettivo è duplice: sviluppare metodi statistici robusti motivati da un solido background teorico e proporre algoritmi efficienti, scalabili e trattabili per le loro applicazioni. \\ La tesi è organizzata come segue. Nel capitolo 1 esamineremo brevemente il background metodologico e discuteremo i concetti necessari che appartengono alle diverse aree a cui contribuiremo con questa tesi. \\ Nel capitolo 2 proponiamo un modello di atomi comuni (CAM) per nested data, che supera le limitazioni del processo del nested Dirichlet Process, come discusso in \ citep {Camerlenghi2018}. Deriviamo le sue proprietà teoriche e sviluppiamo uno slice sampler per dati nested al fine di ottenere un algoritmo efficiente per la simulazione della posterior. Abbiamo poi incorporato il modello in un framework di Rounded mixture of Gaussian Kernels, così da applicare il nostro metodo a una abundance table derivante da uno studio di microbioma. \\ Nel capitolo \ref {BNPT} sviluppiamo una versione BNP del two-group model, modellando sia $ f_0 $ che $ f_1 $ con Pitman-Yor mixtures models. Proponiamo di fissare i due parametri $ \sigma_0 $ e $ \sigma_1 $ in modo che $ \sigma_0> \sigma_1 $, in base alla logica secondo cui il PY che modella la distribuzione nulla dovrebbe essere più vicino alla sua misura di base (opportunamente scelta Gaussiana standard), mentre il PY alternativo dovrebbe avere meno vincoli. Per indurre la separazione, impieghiamo una non-local prior sul parametro location della misura base del PY collocato su $ f_1 $. Mostriamo come il modello si comporta in diversi scenari e applichiamo questa metodologia a un set di dati del microbioma. \\ Il capitolo \ref{Peluso} presenta una seconda proposta per il two-group model. Qui, utilizziamo non-local distributions per modellare la densità alternativa direttamente nella formulazione della Likelihood. Abbiamo proposto una formulazione sia parametrica che non parametrica del modello. Forniamo poi una giustificazione teorica per l'adozione di questo approccio e, dopo aver confrontato le prestazioni del nostro modello con diversi concorrenti, presentiamo tre applicazioni su set di dati genomici reali pubblicamente disponibili. \\ Nel capitolo \ref {CRIME} ci concentriamo sul miglioramento del modello per la stima delle dimensioni intrinseche (ID) discusso in \citet {Allegra}, dove gli autori stimano gli IDs modellando il rapporto delle distanze da un punto dal suo primo e secondo vicino più vicino (NN). Innanzitutto, proponiamo di includere distribuzioni a priori più adatte nel loro modello mistura finita. Quindi, estendiamo la metodologia teorica esistente derivando distribuzioni in forma chiusa per i rapporti di distanze da un punto a due NNs di ordine generico. Proponiamo poi un semplice modello di mistura nonparametrica usando il processo di Dirichlet, in cui sfruttiamo le distribuzioni derivate per estrarre più informazioni dai dati. Il capitolo si conclude quindi con studi di simulazione e l'applicazione a dati reali. \\ Infine, il capitolo \ref {Conclusions} presenta le direzioni future e le conclusioni. Bayesian mixture models are ubiquitous in statistics due to their simplicity and flexibility and can be easily employed in a wide variety of contexts. In this dissertation, we aim at providing a few contributions to current Bayesian data analysis methods, often motivated by research questions from biological applications. In particular, we focus on the development of novel Bayesian mixture models, typically in a nonparametric setting, to improve and extend active research areas that involve large-scale data: the modeling of nested data, multiple hypothesis testing, and dimensionality reduction.\\ Therefore, our goal is twofold: to develop robust statistical methods motivated by a solid theoretical background, and to propose efficient, scalable and tractable algorithms for their applications.\\ The thesis is organized as follows. In Chapter \ref{intro} we shortly review the methodological background and discuss the necessary concepts that belong to the different areas that we will contribute to with this dissertation. \\ In Chapter \ref{CAM} we propose a Common Atoms model (CAM) for nested datasets, which overcomes the limitations of the nested Dirichlet Process, as discussed in \citep{Camerlenghi2018}. We derive its theoretical properties and develop a slice sampler for nested data to obtain an efficient algorithm for posterior simulation. We then embed the model in a Rounded Mixture of Gaussian kernels framework to apply our method to an abundance table from a microbiome study.\\ In Chapter \ref{BNPT} we develop a BNP version of the two-group model \citep{Efron2004}, modeling both the null density $f_0$ and the alternative density $f_1$ with Pitman-Yor process mixture models. We propose to fix the two discount parameters $\sigma_0$ and $\sigma_1$ so that $\sigma_0>\sigma_1$, according to the rationale that the null PY should be closer to its base measure (appropriately chosen to be a standard Gaussian base measure), while the alternative PY should have fewer constraints. To induce separation, we employ a non-local prior \citep{Johnson} on the location parameter of the base measure of the PY placed on $f_1$. We show how the model performs in different scenarios and apply this methodology to a microbiome dataset.\\ Chapter \ref{Peluso} presents a second proposal for the two-group model. Here, we make use of non-local distributions to model the alternative density directly in the likelihood formulation. We propose both a parametric and a nonparametric formulation of the model. We provide a theoretical justification for the adoption of this approach and, after comparing the performance of our model with several competitors, we present three applications on real, publicly available genomic datasets.\\ In Chapter \ref{CRIME} we focus on improving the model for intrinsic dimensions (IDs) estimation discussed in \citet{Allegra}. In particular, the authors estimate the IDs modeling the ratio of the distances from a point to its first and second nearest neighbors (NNs). First, we propose to include more suitable priors in their parametric, finite mixture model. Then, we extend the existing theoretical methodology by deriving closed-form distributions for the ratios of distances from a point to two NNs of generic order. We propose a simple Dirichlet process mixture model, where we exploit the novel theoretical results to extract more information from the data. The chapter is then concluded with simulation studies and the application to real data.\\ Finally, Chapter \ref{Conclusions} presents the future directions and concludes.

This research defines the public response to changes towards low carbon energy systems. The findings support regional and local energy planners in order to accelerate the energy transition in the Alpine local communities, compatibly with their resources and specificities. Energy transition is a long-term change process of the energy system towards low carbon features (i.e., more energy production from renewable sources, higher energy efficiency). From a technological point of view, energy system is an integrated system including several energy sources, technologies, and products for energy production, distribution, and consumption. From a social point of view, local population's choices and actions determine the time and the features of energy transition at regional and local scales. Indeed, energy system is not only technology matter. Socio-energy and territorial approaches underline the importance of features and relationships between energy, society, and space. Based on the interactions between energy, society, and space, this research delineates an analytical framework and an applied definition of socio-energy system at local and translocal scale. Through a systematic literature review of 168 scientific publications, this research defines the list of dimensions and key aspects that shape an energy system at regional and local scales by means of local population's choices. This research uses these dimensions to analyse the socio-energy systems and their variation within a regional territory. The transformation of these dimensions into variables is the basis for the cluster analysis applied to the South Tyrol case study (Italy). This analysis defines eight different socio-energy systems within this regional territory. Each socio-energy system has its own specificities and resources to take into consideration in the energy planning for accelerating the energy transition. With the aim to propose recommendations on how to use social and territorial specificities as levers for wider achievement of energy targets at local scale, this research proposes a Decision Support Tool addressed to regional and local energy planners. The Decision Support Tool also reports potential collaborations between municipalities with similar specificities and needs in order to save, protect, and share resources for a more effective energy planning. The collaborations proposed by the Decision Support Tool only partially coincide with the actual collaborations in the energy governance of South Tyrol. Through a Bayesian exponential random graph model, this research identifies the network structures and dynamics that are at the basis of the actual energy governance. Further, the analysis identifies some network configurations that might change the actual network towards an energy governance that consider the local and translocal specificities. Concluding, the methodologies used in this research are replicable to other case studies with small adjustments.

In this work we present a novel statistical approach to improve the assessment of the adherence to a 1-year nutritional intervention within the framework of the NU-AGE project. This was measured with a single adherence score based on 7-days food records, under limitations on the number of observations per subject and time frame of intervention. The results of the NU-AGE dietary intervention were summarized by variations of the NU-AGE index as described in the NU-AGE protocol. Food and nutrient intake of all participants was assessed by means of 7-days food records at recruitment and after 10 to 14 months of intervention (depending on the subject availability). Sixteen food groups and supplementations covering the dietary goals of the NU-AGE diet have been used to estimate the NU-AGE index before and after the intervention. The 7-days food record is a reliable tool to register food intakes, however, as with other tools used to assess lifestyle dietary compliance, it is affected by uncertainty in this estimation due to the possibility that the observed week is not fully representative of the entire intervention period. Also, due to logistic limitations, the effects of seasonality can never be completely removed. These variabilities, if not accounted for in the index estimation, will reduce the statistical power of the analyses. In this work we discuss a method to assess these uncertainties and thus improve the resulting NU-AGE index. The proposed method is based on Hierarchical Bayesian Models. This model explicitly includes country-specific averages of the NU-AGE index, index variation induced by the dietary intervention, and country based seasonality. This information is used to evaluate the NU-AGE index uncertainty and thus to estimate the "real" NU-AGE index for each subject, both before and after the intervention. These corrections reduce the possibility of misinterpreting measurement variability as real information, improving the power of the statistical tests that are performed with the resulting index. The results suggest that this method is able to reduce the short term and seasonal variability of the measured index in the context of multicenter dietary intervention trials. Using this method to estimate seasonality and variability would allow one to obtain better measurements from the subjects of a study, and be able to simplify the scheduling of diet assessments. Clinical Trial Registration: www.ClinicalTrials.gov, identifier NCT01754012.

The standard version of the game Rock-Paper-Scissors is interesting in terms of game theory, but less so in terms of Statistics. However, we show that with a small rule change it can be made into an interactive exercise for degree-level students of Statistics that leads to a Bayesian change-point model, for which the Gibbs sampler provides an intuitive method of inference. First, students play the game to generate the data. Second, they are encouraged to formulate a model that reflects their experience from having played the game. And third, they participate in the development of a suitable MCMC algorithm to fit the model.

The purpose of this study is to infer evolutionary trees through the Markovchain Monte Carlo algorithm (MCMC) [1] based on whole-genome single cell DNA sequencing data. By using MCMC we obtain likely tree structure samples according to the cells’ somatic point mutations in our data. This probabilistic framework takes into consideration the errors caused by the current technology such as amplification errors, sequencing errors and allelic dropouts. We investigated whether using this technique is reasonable given this biological scope. Most of the results give interesting conclusions that improve the previous results on the same Site Pair Model [2] and therefore we conclude that using MCMC is reasonable. Though, since the model is based on probabilities and the algorithm randomizes decisions the best results are not always guaranteed. One needs to be aware that a decent amount of data in the dataset is an important requisite to predict accurate tree structures. Furthermore, the computational time for this process is significantly high and can not be computed on regular laptops for large and realistic data sets. This is acceptable since for this type of research speed is not a strict requirement: it is worth waiting more for a given execution if the obtained results are more interestingor more accurate. Finally, we propose some further improvements for this strategy that could potentially obtain even better results in terms of accuracy and speed. Syftet med denna studie är att dra slutsatser om evolutionära träd genom Markov-kedjan Monte Carlo-algoritmen (MCMC) [1] baserat på DNA sekvenseringsdata med hela genomet. Genom att använda MCMC får vi sannolikt trädstruktur prover enligt cellenssomatiska punktmutationer i våra data. Detta probabilistiska ramverk tar hänsyn till de fel som orsakas av den nuvarande teknologin, såsom förstärkningsfel, sekvenseringsfel och alleliska utfall. Vi undersökte om användningen av denna teknik är rimlig med tanke på detta biologiska omfång. De flesta av resultaten ger intressanta slutsatser som förbättrar tidigare resultat på samma Site Pair Model [2] och drar därför slutsatsen att MCMC är rimligt. Då modellen är baserad på sannolikheter och algoritmen randomiserar beslut är de bästa resultaten inte alltid garanterade. Man måste vara medveten om att användningen av en tillbördigmängd data i datasatsen är en viktig förutsättning för att förutsäga exakta trädstrukturer. Dessutom är beräkningstiden för denna process betydligt hög och kan inte beräknas på vanliga bärbara datorer för stora och realistiskadataset. Detta är acceptabelt eftersom för denna typ av forskning är inte hastighet ett strikt krav: det är värt att vänta längre på ett visst utförande om de erhållna resultaten är mer intressanta eller mer exakta. Slutligen föreslår vi några ytterligare förbättringar för denna strategi som potentiellt kan få ännu bättre resultat när det gäller noggrannhet och snabbhet.

Mugil liza Valenciennes, 1836 is an economically important food fish and has been recommended for aquaculture in South America. A total of 278 fishes were collected in the spring and summer of 2009 and 2010. These fish were sorted into sample groups according to their size class. We used Bayesian statistics and 95% credible intervals for each parameter tested were calculated. Fish studied harbored a total of 15 different species of parasites. Diversity of parasite species found on Mugilliza was greatest at the S.R.C. collection site, but evidenced a lower species richness than at A.R. site. The 1st size fishes of both sites evidenced greater parasite diversity than either 2nd or 3rd size fish. Differences observed could be explained by the different use of habitat types at the two sites or differential susceptibility to infection by parasites. The dominance of D. fastigata influenced observed results of lower community diversity indexes. New works elucidating different parasite life cycles within juvenile and adults of M. liza in Argentina, promise to be important for determining the risk of the parasitism by zoonotic metacercariae A. (P.) longa and use of this fish as food and an economic resource, and the possible use of mullet parasites in other promising fields as indicators of biodiversity, and/ or water contamination., Mugil liza, Valenciennes, 1836 es un pez de importancia comercial y ha sido recomendado en Sudamerica para la acuicultura. Los peces se distribuyeron en grupos de acuerdo con la talla. Usamos estadística Bayesiana e intervalos de credibilidad del 95% para cada parámetro calculado. Los peces alojaron un total de 15 especies parasitarias. La diversidad de las especies parásitas de M. liza fue mayor en el sitio de colecta S.R.C., pero evidencio una menor riqueza especifica que en A.R. La 1er talla de los peces de ambos sitios evidencio una mayor diversidad parasitaria que la 2da o 3er talla. Las diferencias observadas se pueden explicar por un uso diferente del hábitat en los dos sitios o a una susceptibilidad diferente a la infección por los parásitos. La influencia de la dominancia de D. fastigata resulta en el menor índice de diversidad de la comunidad. Nuevos trabajos que diluciden los diferentes ciclos de vida de los parásitos en los juveniles y adultos de M. liza en Argentina, prometen ser importantes para determinar el riesgo del parasitismo de la metacercaria zoonotica A. (P.) longa y el uso de este pez como alimento, una fuente económica y el posible uso de los parásitos de la lisa en otros promisorios campos como indicadores de biodiversidad y/o contaminación acuática., Centro de Estudios Parasitológicos y de Vectores

En muchas situaciones resulta util cuanticar informacion subjetiva que uno o varios expertos conocen acerca de un tema de interes, por esto, una parte importante dentro de la estadstica Bayesiana es la construccion de metodos de elicitacion para hallar distribuciones de probabilidad. Con el fin de contribuir al desarrollo en este campo,se desarrollo una metodología para elicitar los parametros de la regresion logstica con una sola covariable. El metodo que se plantea requiere que se jen unos niveles de la covarible y en estos se asume una distribucion Binomial, para cada nivel se elicita el parametro de interes mediante la metodologa indirecta de muestras hipoteticas. In many situations it is useful to quantify subjective information that one or more experts know about a topic of interest, therefore, an important part of Bayesian statistics is building elicitation methods for finding probability distributions. In order to contribute to the development in this field, a methodology was developed to elicit the parameters of the logistic regression with a single covariate. The proposed method requires to determite the levels of the covarible, at each level a binomial distribution is assumed and the parameter of interest is elicited by using the indirect method hypothetical samples.