Thesis (Ph. D.)--University of Wisconsin--Madison, 1967. / Typescript. Vita. eContent provider-neutral record in process. Description based on print version record. Includes bibliography.

Thesis (Ph. D.)--Oregon State University, 2007. / Printout. Includes bibliographical references (leaves 130-132). Also available on the World Wide Web.

Η ανάλυση διασποράς μελετά τη σχέση που έχει μια εξαρτημένη μεταβλητή Υ, τις τιμές της οποίας μπορούμε να παρτηρήσουμε, με τις τιμές ενός παράγοντα Α ή πολλών παραγόντων που είναι και οι ανεξάρτητες μεταβλητές. Είναι κατάλληλη για δεδομένα που βασίζονται πάνω στην έρευνα ή σε πειράματα και μπορεί να χρησιμοποιηθεί είτε οι ανεξάρτητες μεταβλητές είναι ποσοτικές είτε ποιοτικές. Στην παρούσα εργασία, θα χρησιμοποιήσουμε ένα μοντέλο ανάλυσης διασποράς για να μελετήσουμε τις επιρροές των ανεξάρτητων μεταβλητών στην εξαρτημένη μεταβλητή χωρίς περιορισμούς στην φύση της στατιστικής σχέσης καθώς επίσης, θα γίνουν έλεγχοι για να μελετήσουμε αν οι πληθυσμοί μας έχουν ίσες μέσες τιμές ή ίσες διασπορές. Στο πρώτο κεφάλαιο της παρούσας εργασίας παρουσιάζονται τα γενικά χαρακτηριστικά μιας μονοπαραγοντικής ανάλυσης διασποράς καθώς επιίσης και οι έλεγχοι των μέσων τιμών για τις στάθμες του παράγοντα. Στο δεύτερο κεφάλαιο ακολουθεί η παρουσίαση της ανάλυσης των επιδράσεων του παράγοντα και οι διάφορες μέθοδοι για πολλαπλές συγκρίσεις. Επίσης, στο τρίτο κεφάλαιο αναλύεται ο σχεδιασμός των δειγματικών μεγεθών καθώς και οι έλεγχοι για ίσες διασπορές. Τέλος, στο τέταρτο κεφάλαιο αναφέρονται τα γενικά χαρακτηριστικά μιας πολυπαραγοντικής ανάλυσης διασποράς. / The analysis of variance studies the relation which has a dependent variable Y, whose the values we observe, with the values of a factor A or more factors that are the independent variables. It is suitable for data that are based on a research or on experiments and it can be used either the independent variables are quantitative or qualitative. In this paper, we are going to use a model of analysis of variance to study the influences of the independent variables to the dependent variable without restrictions to the nature of the statistical relation and they will be controls to study if our populations have equal means or variances, as well. In the first chapter of this paper, they are introduced the general characteristics of a single-factor analysis of variance and the controls of the means for the levels of the factor, as well. In the second chapter, it follows the presentation of the analysis of the factor effects and several methods for multiple comparisons. Furthermore, in the third chapter, it is analysed the planning of sample sizes and the controls for equal variances, as well. In the last chapter, the general charachteristics of a multiple analysis of variance are reported.

Background: A key driver for early years strategies is the reduction of oppositional and defiant behaviour in childhood to prevent a negative life course of poor educational attainment and criminality. Despite a robust evidence base, manualised parent skills training programmes (PT) for externalising behaviour problems are only effective for approximately two-thirds of families. A limited number of variables that account for variance in outcome have been discovered. Finding further predictor, moderator and mediator variables will explain who benefits most, and how change occurs. This will ensure that families receive the most appropriate treatments for their profile of needs, and services deliver the available interventions in an efficient and effective way. Objectives: A systematic review of the literature was conducted to explore progress in this area since two key meta-analyses published in 2006. A primary study was carried out to examine whether parent attachment style, parenting self-efficacy and dysfunctional parental attributions predict, moderate or mediate the levels of externalising child behaviour problems reported by parents attending the Incredible Years PT. Methods: Studies exploring variables influencing outcome in child behaviour following attendance at a manualised, evidence-based PT group for parents of children and adolescents aged 0-18 years were sought. Psychinfo, Medline, ERIC and Embase databases were searched for articles published between August 2004 and March 2013 with keywords ‘parent’, ‘child’, ‘training’, ‘indirect effects’ and ‘oppositional behaviour’ or related terms. 2853 articles were retrieved, from which 12 studies fulfilled criteria. Study quality was appraised and co-rated. A pre-post, within subjects design was conducted with 79 parents attending the Incredible Years PT delivered in a Child and Adolescent Mental Health Service. Participants completed a battery of pre-treatment questionnaires measuring attachment style, attributions, self-efficacy and child behaviour. 52 parents completed the same battery post-treatment, and missing data was carried forward in an intent to treat analysis. Data was analysed using multiple regression techniques, and mediation and moderated mediation analyses. Results: The recent evidence base is populated by secondary analyses of intervention RCTs, and less robust non RCTs. The selection of maternal mood, parenting stress, parenting style and child demographics dominate, and the exploration of unique variables is limited. Significant findings are mixed and add no new variables to our understanding. Significant changes in parenting self-efficacy and dysfunctional attributions were found post-treatment, and attachment style remained stable. A main treatment effect size of d=.3 was estimated, and a significant number of children fell below sub-clinical levels of problem behaviour (n=13, 15.7%). Baseline child-responsible attributions and self-efficacy accounted for up to 40% of the variance in baseline child behaviour. Attachment style did not contribute significantly to the model, but moderated parent-causal attributions. Post-treatment, attachment avoidance had a moderating effect on self-efficacy and child-responsible attributions, and a significant direct effect on outcome. The indirect effect of parental-attributions on child behaviour through self-efficacy was moderated by attachment avoidance which reduced the number of significant paths. Conclusions: The call for PT studies delivered with fidelity in real world settings has been recognised, and more sophisticated statistical models of mediation are being adopted. There remains an exhaustive list of novel potential variables that future research needs to select and explore in primary research designs. An evidence based PT is achieving statistically and clinically significant results for children referred for problem behaviour. Dysfunctional parent attributions and self-efficacy are predictors of both pre- and post-treatment levels of child behaviour, which could be screened for in the referral process. The evidence for a direct and indirect role of attachment style on parent training outcomes adds a new candidate variable to the literature that warrants further exploration.

[pt] Em algumas aplicações, como por exemplo sinais de dados, a variância do sinal pode ser desconhecida, porém constante. Nesses casos, quantizadores adaptivos que utilizam algoritmos de estimação local da variância não são apropriados para a discretização do sinal. Algoritmos mais adequados para essa situação são aqueles que se preocupam em aprender a variância do sinal de entrada. Neste trabalho são examinados quatro algoritmos de aprendizagem de variância, com vistas ao seu emprego em quatização adaptiva. Um destes algoritmos, proposto por A. Gersho e D. J. Goodman, é um algoritmo de aproximação estocástica que converge com probabilidade 1. É mostrado que um outro algoritmo, também de aproximação estocástica, converge com probabilidade 1 para a aplicação em um quantizador adaptivo com entradas independentes. Os outros dois algoritmos consistem de modificações introduzidas sobre dois primeiros, com a finalidade de obter uma maior velocidade de convergência. Finalmente, é analisado, através de simulações em computador, o desempenho desses quatro quantizadores adaptivos quando usados em sistemas DPCM. / [en] In some applications, such as data transmission, signal variance is unknown but constant. In such cases, adaptive quantizers using local variance estimation algorithms are not appropriate for the signal quantizations. The most suitable algorithms for this situation are those which learn the input signal variance. This work examines four variance learning algorithms for application in adaptive quantization. One of them, proposed by A. Gersho and D. J. Goodman, is a stochastic approximation algorithm which converges with probability one, when applied to adaptive quantization. The remaining two algorithms are modified versions of the first two, in order to obtain greater convergence speed. Finally, performance of these four adaptive quantizers, when used in DPCM systems, is analyzed through computer simulations.

Cette thèse examine la dispersion multivariée des modelés normales stables Tweedie. Trois estimateurs de fonction variance généralisée sont discutés. Ensuite dans le cadre de la famille exponentielle naturelle deux caractérisations du modèle normal-Poisson, qui est un cas particulier de modèles normales stables Tweedie avec composante discrète, sont indiquées : d'abord par fonction variance et ensuite par fonction variance généralisée. Le dernier fournit la solution à un problème particulier d'équation de Monge-Ampère. Enfin, pour illustrer l'application de la variance généralisée des modèles Tweedie stables normales, des exemples à partir des données réelles sont fournis. / This thesis examines the multivariate dispersion of normal stable Tweedie (NST) models. Three generalize variance estimators of some NST models are discussed. Then within the framework of natural exponential family, two characterizations of normal Poisson model, which is a special case of NST models with discrete component, are shown : first by variance function and then by generalized variance function. The latter provides a solution to a particular Monge-Ampere equation problem. Finally, to illustrate the application of generalized variance of normal stable Tweedie models, examples from real data are provided.

This study sought to determine the effect of varying degrees of disproportionality of four methods of handling disproportionality cell frequencies in two-way analysis of variance. A Monte Carlo simulation procedure was employed. Two multiple linear regression techniques and two "approximate" techniques were compared.

The purpose of this study is to investigate the pricing of variable annuity embedded derivatives in a Lévy process setting. This is one of the practical issues that continues to face life insurers in the management of derivatives embedded within these products. It also addresses how such providers can protect themselves against adverse scenarios through a hedging framework built from the pricing framework. The aim is to comparatively consider the price differentials of a life insurer that prices its variable annuity guarantees under the more actuarially accepted regime-switching framework versus the use of a Lévy framework. The framework should address the inadequacies of conventional deterministic pricing approaches used by life insurers given the increasing complexity of the option-like products sold. The study applies finance models in the insurance context given the similarities in payoff structure of the products offered while taking into account the differences that may exist. The underlying Lévy process used in this study is the Variance-Gamma (VG) process. This process is useful in option pricing given its ability to model higher moments, skewness and kurtosis, and also incorporate stochastic volatility. The research results compare well with the regime-switching framework besides the added merit in the use of a more refined model for the underlying that captures most of the observed market dynamics. / Dissertation (MSc)--University of Pretoria, 2014. / tm2015 / Mathematics and Applied Mathematics / MSc / Unrestricted

L’analyse de mesures longitudinales –appelées trajectoires– est de plus en plus fréquente en recherche médicale. L’un des intérêts de cette analyse est d’identifier des groupes d’individus ayant des trajectoires similaires. La classification obtenue peut être utilisée pour mieux comprendre l’hétérogénéité des évolutions entre individus. La classification peut être déterminée à partir d’un modèle pour lequel les trajectoires des individus correspondent à la trajectoire du groupe auquel ils sont affectés. L’objectif de la thèse est de développer une extension de ce modèle de classification standard permettant une meilleure prise en compte de la variabilité au sein des groupes, (i) variabilité des valeurs du marqueur (variance résiduelle) et (ii) variabilité des profils d’évolution (variance inter-individuelle). Deux modèles de classification sont développés : 1) un premier modèle qui prend en compte une variance résiduelle au sein de chaque groupe variable d’un groupe à l’autre, et 2) un deuxième modèle qui prend en compte une variabilité des trajectoires au sein des groupes au lieu de de prédire la même trajectoire pour tous les individus d’un même groupe, variabilité qui peut être identique ou variable d’un groupe à l’autre. L’intérêt de ces deux modèles a été montré par des travaux de simulations et par des applications cliniques. Globalement, lorsque le nombre de mesures et de trajectoires est suffisant, ces modèles donnent de meilleures classifications que celles du modèle de classification standard. Par ailleurs, en dehors de plans expérimentaux très contrôlés, les deux sources de variabilité sont inhérentes à la recherche en santé. Ces modèles sont donc très pertinents d’un point de vue clinique / The analysis of longitudinal measures –called trajectories– is more and more frequent in clinical research. One of the interests of this analysis is to identify groups of individuals with similar trajectories. The obtained classification is used to understand and explore the heterogeneity of trajectories among subjects. The classification can be performed by a model that predicts the same trajectory for all the subjects that are classified in the same group. The objective of this thesis is to develop an extension to the standard classification model that gives greater consideration to the variability within groups, (i) the variability of marker values (residual variance), and (ii) the variability of the individual trajectories inside a group (between-individual variance). Two classification models were developed: 1) a first model that allows unequal residual variance across groups, and 2) a second model that takes into account a between-individual variance within each group instead of predicting the same trajectory for all subjects in the same group, a variance that can be equal or unequal across groups. The interest of these two models has been studied by simulations and through clinical applications. Overall, when the number of trajectories and measurements per trajectory is sufficient, these models gives better classification compared to the standard classification model. Moreover, except for highly controlled experimental designs, the two sources of variability are inherent to research in health. Therefore, these models are very relevant from a clinical point of view.

Cette thèse aborde le problème de l’estimation d’état de systèmes dynamiques à entrées inconnues (EI). La première partie de la thèse est consacrée aux observateurs ponctuels. Pour les systèmes linéaires à temps variant (LTV) ou à paramètres variants (LPV) avec EI et bruits gaussiens, un nouvel observateur à minimum de variance est proposé : il affiche des résultats équivalents aux observateurs existants en matière de performance, mais les surpasse par la simplicité de son implémentation et par l’existence d’un résultat de stabilité qui lui est associé. Pour les systèmes linéaires à commutations avec EI, bruits gaussiens et séquence de commutation inconnue, un premier estimateur à minimum de variance est proposé. Pour les systèmes non linéaires, un nouveau filtre de Kalman étendu à EI est proposé: il généralise les résultats existants au cas de systèmes dont l’équation de mesure possède une partie linéaire et une partie non linéaire. Une application à la navigation aérospatiale est proposée. Pour les systèmes LTV et LPV à EI et pertubations à énergie finie, un observateur H∞ est proposé : il généralise et unifie certains résultats existants. La deuxième partie de la thèse est consacrée aux estimations garanties. Dans lecas où les perturbations affectant le système dynamique sont bornées, plusieurs observateurs par intervalle sont construits. Pour des systèmes LPV avec EI, le travail généralise l’existant en relaxant une condition de rang largement utilisée pour le découplage de l’EI. Pour les systèmes linéaires à commutations sans ou avec EI, un observateur par intervalle est proposé pour traiter le cas où la séquence de commutation est inconnue, problème qui n’a jamais été traité jusqu’ici. Finalement, une nouvelle approche d’encadrement de l’état par un observateur algébrique est proposée. / This thesis deals with the problem of state estimation for dynamical systems affected by Unknown Inputs (UI).The first part focuses on point-wise observers. For Linear Time Varying (LTV) or Linear Parameter Varying (LPV) systems with UI and Gaussian noises, a new minimum variance observer is proposed. It equals the existing ones in terms of performance, but overpasses them by its simplicity of implementing and by the existence of an associated stability result. For linear switching systems with Gaussian noises, UI and unknown switching sequence, a first minimum variance estimator is provided. For non linear systems, a new UI Extended Kalman Filter is provided: it generalizes existing results to the case the measurement equation possesses both linear and nonlinear parts. An application for aerospace navigation is proposed. For LTV or LPV systems with UI and finite energy perturbations, an H∞ observer is given : it generalizes and unifies some existing results. The second part focuses on guaranteed set observers. If the perturbations affecting the dynamical system are bounded,several interval observers are provided. For LPV systems with UI, it generalizes some existing results by relaxing a rank condition widely used for UI decoupling. For linear switching systems without or with UI, it provides an interval observer in the case the switching sequence is unknown, which has not been done before. Finally, a new algebraic observer for interval estimation is proposed.