Improved numerical reservoir models are constructed when all available diverse data sources are accounted for to the maximum extent possible. Integrating various diverse data is not a simple problem because data show different precision and relevance to the primary variables being modeled, nonlinear relations and different qualities. Previous approaches rely on a strong Gaussian assumption or the combination of the source-specific probabilities that are individually calibrated from each data source. This dissertation develops different approaches to integrate diverse earth science data. First approach is based on combining probability. Each of diverse data is calibrated to generate individual conditional probabilities, and they are combined by a combination model. Some existing models are reviewed and a combination model is proposed with a new weighting scheme. Weakness of the probability combination schemes (PCS) is addressed. Alternative to the PCS, this dissertation develops a multivariate analysis technique. The method models the multivariate distributions without a parametric distribution assumption and without ad-hoc probability combination procedures. The method accounts for nonlinear features and different types of the data. Once the multivariate distribution is modeled, the marginal distribution constraints are evaluated. A sequential iteration algorithm is proposed for the evaluation. The algorithm compares the extracted marginal distributions from the modeled multivariate distribution with the known marginal distributions and corrects the multivariate distribution. Ultimately, the corrected distribution satisfies all axioms of probability distribution functions as well as the complex features among the given data. The methodology is applied to several applications including: (1) integration of continuous data for a categorical attribute modeling, (2) integration of continuous and a discrete geologic map for categorical attribute modeling, (3) integration of continuous data for a continuous attribute modeling. Results are evaluated based on the defined criteria such as the fairness of the estimated probability or probability distribution and reasonable reproduction of input statistics. / Mining Engineering

Generating random variates as generalisation of a given sample is an important task for stochastic simulations. The three main methods suggested in the literature are: fitting a standard distribution, constructing an empirical distribution that approximates the cumulative distribution function and generating variates from the kernel density estimate of the data. The last method is practically unknown in the simulation literature although it is as simple as the other two methods. The comparison of the theoretical performance of the methods and the results of three small simulation studies show that a variance corrected version of kernel density estimation performs best and should be used for generating variates directly from a sample. (author's abstract) / Series: Preprint Series / Department of Applied Statistics and Data Processing

This dissertation investigates three topics related to imagereconstruction from list-mode Anger camera data. Our mainfocus is the processing of photomultiplier-tube (PMT)signals directly into images. First we look at the use of list-mode calibration data toreconstruct a non-parametric likelihood model relating theobject to the data list. The reconstructed model can thenbe combined with list-mode object data to produce amaximum-likelihood (ML) reconstruction, an approach we calldouble list-mode reconstruction. This trades off reducedprior assumptions about the properties of the imaging systemfor greatly increased processing time and increaseduncertainty in the reconstruction. Second we use the list-mode expectation-maximization (EM)algorithm to reconstruct planar projection images directlyfrom PMT data. Images reconstructed by EM are compared withimages produced using the faster and more common techniqueof first producing ML position estimates, then histogramingto form an image. A mathematical model of the human visualsystem, the channelized Hotelling observer, is used tocompare the reconstructions by performing the Rayleigh task,a traditional measure of resolution. EM is found to producehigher resolution images than the histogram approach,suggesting that information is lost during the positionestimation step. Finally we investigate which linear parameters of an objectare estimable, in other words may be estimated without biasfrom list-mode data. We extend the notion of a linearsystem operator, familiar from binned-mode systems, tolist-mode systems, and show the estimable parameters aredetermined by the range of the adjoint of the systemoperator. As in the binned-mode case, the list-modesensitivity functions define ``natural pixels'' with whichto reconstruct the object.

<p>In this thesis, we develop some Bayesian mixture density estimation for univariate and multivariate data. We start proposing a repulsive process favoring mixture components further apart. While conducting inferences on the cluster-specific parameters, current frequentist and Bayesian methods often encounter problems when clusters are placed too close together to be scientifically meaningful. Current Bayesian practice generates component-specific parameters independently from a common prior, which tends to favor similar components and often leads to substantial probability assigned to redundant components that are not needed to fit the data. As an alternative, we propose to generate components from a repulsive process, which leads to fewer, better separated and more interpretable clusters. </p><p>In the second part of the thesis, we face the problem of modeling the conditional distribution of a response variable given a high dimensional vector of predictors potentially concentrated near a lower dimensional subspace or manifold. In many settings it is important to allow not only the mean but also the variance and shape of the response density to change flexibly with features, which are massive-dimensional. We propose a multiresolution model that scales efficiently to massive numbers of features, and can be implemented efficiently with slice sampling.</p><p> In the third part of the thesis, we deal with the problem of characterizing the conditional density of a multivariate vector of response given a potentially high dimensional vector of predictors. The proposed model flexibly characterizes the density of the response variable by hierarchically coupling a collection of factor models, each one defined on a different scale of resolution. As it is illustrated in Chapter 4, our proposed method achieves good predictive performance compared to competitive models while efficiently scaling to high dimensional predictors.</p> / Dissertation

This thesis examines a newly proposed density estimator in order to evaluate its usefulness for government crop insurance programs confronted by the problem of adverse selection. While the Federal Crop Insurance Corporation (FCIC) offers multiple insurance programs including Group Risk Plan (GRP), what is needed is a more accurate method of estimating actuarially fair premium rates in order to eliminate adverse selection. The Empirical Bayes Nonparametric Kernel Density Estimator (EBNKDE) showed a substantial efficiency gain in estimating crop yield densities. The objective of this research was to apply EBNKDE empirically by means of a simulated game wherein I assumed the role of a private insurance company in order to test for profit gains from the greater efficiency and accuracy promised by using EBNKDE. Employing EBNKDE as well as parametric and nonparametric methods, premium insurance rates for 97 Illinois counties for the years 1991 to 2010 were estimated using corn yield data from 1955 to 2010 taken from the National Agricultural Statistics Service (NASS). The results of this research revealed substantial efficiency gain from using EBNKDE as opposed to other estimators such as Normal, Weibull, and Kernel Density Estimator (KDE). Still, further research using other crops yield data from other states will provide greater insight into EBNKDE and its performance in other situations.

No description available.

Un défi majeur dans les méthodes d'estimation non linéaire multi-échelle, comme le seuillage des ondelettes, c'est l'extension de ces méthodes vers une disposition où les observations sont irrégulières et non équidistantes. L'application de ces techniques dans le lissage de données ou l'estimation des fonctions de densité, il est crucial de travailler dans un espace des fonctions qui impose un certain degré de régularité. Nous suivons donc une approche différente, en utilisant le soi-disant système de levage. Afin de combiner la régularité et le bon conditionnement numérique, nous adoptons un schéma similaire à la pyramide Laplacienne, qui peut être considérée comme une transformation d'ondelettes légèrement redondantes. Alors que le schéma de levage classique repose sur l'interpolation comme opération de base, ce schéma permet d'utiliser le lissage, en utilisant par exemple des polynômes locaux. Le noyau de l'opération de lissage est choisi de manière multi-échelle. Le premier chapitre de ce projet consiste sur le développement de La transformée polynomiale locale multi-échelle, qui combine les avantages du lissage polynomial local avec la parcimonie de la décomposition multi-échelle. La contribution de cette partie est double. Tout d'abord, il se concentre sur les largeurs de bande utilisées tout au long de la transformée. Ces largeurs de bande fonctionnent comme des échelles contrôlées par l'utilisateur dans une analyse multi-échelle, ce qui s'explique par un intérêt particulier dans le cas des données non-équidistantes. Cette partie présente à la fois une sélection de bande passante optimale basée sur la vraisemblance et une approche heuristique rapide. La deuxième contribution consiste sur la combinaison du lissage polynomial local avec les préfiltres orthogonaux dans le but de diminuer la variance de la reconstruction. Dans le deuxième chapitre, le projet porte sur l'estimation des fonctions de densité à travers la transformée polynomiale locale multi-échelle, en proposant une reconstruction plus avancée, appelée reconstruction pondérée pour contrôler la propagation de la variance. Dans le dernier chapitre, On s’intéresse à l’extension de la transformée polynomiale locale multi-échelle dans le cas bivarié, tout en énumérant quelques avantages qu'on peut exploiter de cette transformée (la parcimonie, pas de triangulations), comparant à la transformée en ondelette classique en deux dimension. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished

In Michelson interferometry, many algorithms are used to detect the number of active laser sources at any given time. Conventional FFT-based non-parametric methods are widely used for this purpose. However, non-parametric methods are not the only possible option to distinguish the peaks in a spectrum, as these methods are not the most suitable methods for short data records and for closely spaced wavelengths. This thesis aims to provide solutions to these problems. It puts forward the use of parametric methods such as autoregressive methods and harmonic methods, and proposes two new algorithms to detect the closely spaced peaks for different scenarios of optical signals in wavemeters. Various parametric algorithms are studied, and their performances are compared with non-parametric algorithms for different criteria, e.g. absolute levels, frequency resolution, and accuracy of peak positions. Simulations are performed on synthetic signals produced from specifications provided by our sponsor, i.e., a wavemeter manufacturing company.

The Malliavin-Thalmaier formula was introduced for simulation of high dimensional probability density functions. But when this integration by parts formula is applied directly in computer simulations, we show that it is unstable. We propose an approximation to the Malliavin-Thalmaier formula. In this paper, we find the order of the bias and the variance of the approximation error. And we obtain an explicit Malliavin-Thalmaier formula for the calculation of Greeks in finance. The weights obtained are free from the curse of dimensionality.

Recent advances in computing technology allow for collecting vast amount of data that arrive continuously in the form of streams. Mining data streams is challenged by the speed and volume of the arriving data. Furthermore, the underlying distribution of the data changes over the time in unpredicted scenarios. To reduce the computational cost, data streams are often studied in forms of condensed representation, e.g., Probability Density Function (PDF). This thesis aims at developing an online density estimator that builds a model called KDE-Track for characterizing the dynamic density of the data streams. KDE-Track estimates the PDF of the stream at a set of resampling points and uses interpolation to estimate the density at any given point. To reduce the interpolation error and computational complexity, we introduce adaptive resampling where more/less resampling points are used in high/low curved regions of the PDF. The PDF values at the resampling points are updated online to provide up-to-date model of the data stream. Comparing with other existing online density estimators, KDE-Track is often more accurate (as reflected by smaller error values) and more computationally efficient (as reflected by shorter running time). The anytime available PDF estimated by KDE-Track can be applied for visualizing the dynamic density of data streams, outlier detection and change detection in data streams. In this thesis work, the first application is to visualize the taxi traffic volume in New York city. Utilizing KDE-Track allows for visualizing and monitoring the traffic flow on real time without extra overhead and provides insight analysis of the pick up demand that can be utilized by service providers to improve service availability. The second application is to detect outliers in data streams from sensor networks based on the estimated PDF. The method detects outliers accurately and outperforms baseline methods designed for detecting and cleaning outliers in sensor data. The third application is to detect changes in data streams. We propose a framework based on Principal Component Analysis (PCA) that reduces the problem of detecting changes in multidimensional data into the problem of detecting changes in the projected data on the principal components. We provide a theoretical analysis, which is support by experimental results to show that utilizing PCA reflects different types of changes in data streams on the projected data over one or more principal components. Our framework is accurate in detecting changes with low computational costs and scales well for high dimensional data.