abstract: It is common in the analysis of data to provide a goodness-of-fit test to assess the performance of a model. In the analysis of contingency tables, goodness-of-fit statistics are frequently employed when modeling social science, educational or psychological data where the interest is often directed at investigating the association among multi-categorical variables. Pearson's chi-squared statistic is well-known in goodness-of-fit testing, but it is sometimes considered to produce an omnibus test as it gives little guidance to the source of poor fit once the null hypothesis is rejected. However, its components can provide powerful directional tests. In this dissertation, orthogonal components are used to develop goodness-of-fit tests for models fit to the counts obtained from the cross-classification of multi-category dependent variables. Ordinal categories are assumed. Orthogonal components defined on marginals are obtained when analyzing multi-dimensional contingency tables through the use of the QR decomposition. A subset of these orthogonal components can be used to construct limited-information tests that allow one to identify the source of lack-of-fit and provide an increase in power compared to Pearson's test. These tests can address the adverse effects presented when data are sparse. The tests rely on the set of first- and second-order marginals jointly, the set of second-order marginals only, and the random forest method, a popular algorithm for modeling large complex data sets. The performance of these tests is compared to the likelihood ratio test as well as to tests based on orthogonal polynomial components. The derived goodness-of-fit tests are evaluated with studies for detecting two- and three-way associations that are not accounted for by a categorical variable factor model with a single latent variable. In addition the tests are used to investigate the case when the model misspecification involves parameter constraints for large and sparse contingency tables. The methodology proposed here is applied to data from the 38th round of the State Survey conducted by the Institute for Public Policy and Michigan State University Social Research (2005) . The results illustrate the use of the proposed techniques in the context of a sparse data set. / Dissertation/Thesis / Ph.D. Mathematics 2011

This study explores the Swedish non-alignment policy from a European Union perspective. Spanning the period 1995-2018, i.e. from the year when Sweden joined the European Union to the present day, the study concerns the way the EU membership has affected Sweden’s defence policy. Given a variety of noted political decisions and developments, mainly in terms of Swedish defence policy propositions during the target period, the key question presents itself as follows: Does the Swedish membership undermine its possibilities to remain as a true case of non-alignment? To be able to answer this question, the study will look into the extent to which the Swedish policy can be seen as a case of Europeanization regarding its EU membership. Thus, the investigation takes as its starting-point the theory of Europeanization. Relying on a variety of previous research, the discussion uses this theory to provide a perspective of this concept as “Goodness of fit”, essentially an analytical tool for identifying processes and changes within this framework. On a practical level, this means that if the different terms and policies fit perfectly between the union and its member states, there is no need for a change. On the other hand, if there is incompatibility in this context, a process of Europeanization is likely to be identified through the lens of “Goodness of fit”. In general terms, it is concluded that the Swedish defence policy has gradually assumed more of a European identity. Arguably, the country’s official non-alignment policy raises questions for its future involvement in the union. Rather, there is indication that political rhetoric has been only one side of the coin, practical reality another. Yet, one thing is clear: it all seems to happen on Swedish terms.

Ce travail porte sur les tests d'adéquation en fiabilité, à la fois pour les systèmes non réparables et les systèmes réparables. Les tests d'adéquation sont des outils efficaces pour vérifier la pertinence d'un modèle pour un jeu de données. Pour les systèmes non réparables, la loi exponentielle et la loi de Weibull sont les lois de durée de vie les plus utilisées en fiabilité. Une comparaison exhaustive des tests d'adéquation pour la loi exponentielle est présentée pour des données complètes et censurées, suivie par des recommandations d'utilisation de ces tests. La loi de Weibull à deux paramètres permet de modéliser des taux de hasard décroissants et croissants contrairement à la loi exponentielle qui suppose un taux de hasard constant. Cependant, il existe moins de tests d'adéquation à la loi de Weibull dans la littérature. Une revue exhaustive des tests existant est effectuée et deux familles de tests exacts sont preésentées. La première famille est la famille des tests basés sur la vraisemblance et la deuxième est la famille des tests basés sur la transformée de Laplace. Des propriétés asymptotiques des nouvelles statistiques de tests sont établies. Une comparaison complète des tests d'adéquation pour la loi de Weibull est effectuée. Des recommandations sur les tests les plus puissants sont données en fonction des caractéristiques du jeu de donnés testé. Pour les systèmes réparables, de nouveaux tests d'adéquation sont développés pour des modèles de maintenance imparfaite avec à la fois des maintenances correctives et des maintenances préventives déterministes. Ces tests sont exacts et peuvent être appliqués à des petits jeux de données. Finalement, des applications à de vrais jeux de données issus de l'industrie sont effectuées pour des systèmes réparables et des systèmes non réparables. / This work deals with goodness-of-fit (GOF) tests in reliability for both non repairable and repairable systems. GOF tests are efficient techniques to check the relevance of a model for a given data set. For non repairable systems, the Exponential and Weibull distributions are the most used lifetimes distributions in reliability. A comprehensive comparison study of the GOF tests for the Exponential distribution is presented for complete and censored samples followed by recommendations about the use of the tests. The two-parameter Weibull distribution allows decreasing and increasing failure rates unlike the Exponential distribution that makes the assumption of a constant hazard rate. Yet, there exist less GOF tests in the literature for the Weibull distribution. A comprehensive review of the existing GOF tests is done and two new families of exact GOF tests are introduced. The first family is the likelihood based GOF tests and the second is the family of tests based on the Laplace transform. Theoretical asymptotic properties of some new tests statistics are established. A comprehensive comparison study of the GOF tests for the Weibull distribution is done. Recommendations about the most powerful tests are given depending on the characteristics of the tested data sets. For repairable systems, new GOF tests are developed for imperfect maintenance models when both corrective maintenance and deterministic preventive maintenance are performed. These tests are exact and can be applied to small data sets. Finally, illustrative applications to real data sets from industry are carried out for repairable and non repairable systems.

La thèse est composée de trois parties. La partie I introduit les outils mathématiques et statistiques appropriés pour l'étude des dépendances, ainsi que des tests statistiques d'adéquation pour des distributions de probabilité empiriques. Je propose deux extensions des tests usuels lorsque de la dépendance est présente dans les données, et lorsque la distribution des observations a des queues larges. Le contenu financier de la thèse commence à la partie II. J'y présente mes travaux concernant les dépendances transversales entre les séries chronologiques de rendements journaliers d'actions, c'est à dire les forces instantanées qui relient plusieurs actions entre elles et les fait se comporter collectivement plutôt qu'individuellement. Une calibration d’un nouveau modèle à facteurs est présentée ici, avec une comparaison à des mesures sur des données réelles. Finalement, la partie III étudie les dépendances temporelles dans des séries chronologiques individuelles, en utilisant les mêmes outils et mesures de corrélations. Nous proposons ici deux contributions à l'étude du « volatility clustering », de son origine et de sa description: l'une est une généralisation du mécanisme de rétro-action ARCH dans lequel les rendements sont auto-excitants, et l'autre est une description plus originale des auto-dépendances en termes de copule. Cette dernière peut être formulée sans modèle et n'est pas spécifique aux données financières. En fait, je montre ici aussi comment les concepts de récurrences, records, répliques et temps d'attente, qui caractérisent la dynamique dans les séries chronologiques, peuvent être écrits dans la cadre unifié des copules. / The thesis is composed of three parts. Part I introduces the mathematical and statistical tools that are relevant for the study of dependences, as well as statistical tests of Goodness-of-fit for empirical probability distributions. I propose two extensions of usual tests when dependence is present in the sample data and when observations have a fat-tailed distribution. The financial content of the thesis starts in Part II. I present there my studies regarding the “cross-sectional” dependences among the time series of daily stock returns, i.e. the instantaneous forces that link several stocks together and make them behave somewhat collectively rather than purely independently. A calibration of a new factor model is presented here, together with a comparison to measurements on real data. Finally, Part III investigates the temporal dependences of single time series, using the same tools and measures of correlation. I propose two contributions to the study of the origin and description of “volatility clustering”: one is a generalization of the ARCH-like feedback construction where the returns are self-exciting, and the other one is a more original description of self-dependences in terms of copulas. The latter can be formulated model-free and is not specific to financial time series. In fact, I also show here how concepts like recurrences, records, aftershocks and waiting times, that characterize the dynamics in a time series can be written in the unifying framework of the copula.

Submitted by Fabio Sobreira Campos da Costa (fabio.sobreira@ufpe.br) on 2016-07-26T12:10:38Z No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) dissertação_final_cd_TT.pdf: 4588819 bytes, checksum: 5127176322bfc06990cbd3eaa1fc5687 (MD5) / Made available in DSpace on 2016-07-26T12:10:38Z (GMT). No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) dissertação_final_cd_TT.pdf: 4588819 bytes, checksum: 5127176322bfc06990cbd3eaa1fc5687 (MD5) Previous issue date: 2016-02-12 / CAPEs / A classe de modelos de regressão beta introduzida por Ferrari & Cribari-Neto (2004) é muito útil para modelar taxas e proporções. O modelo proposto pelos autores é baseado na suposição de que a variável resposta tem distribuição beta com uma parametrização que é indexada pela média e por um parâmetro de precisão. Após a construção de um modelo de regressão é de extrema importância realizar a análise de diagnóstico, objetivando verificar possíveis afastamentos das suposições feitas para o modelo apresentado, bem como detectar possíveis observações que causem influência desproporcional nas estimativas dos parâmetros. A análise de influência local introduzida por Cook (1986) é uma abordagem que objetiva avaliar a influência das observações. Com base no método de influência local, Zhu & Zhang (2004) propuseram um teste de hipóteses para detectar o grau de discrepância entre o modelo suposto e o modelo subjacente do qual dos dados são gerados. Nesse trabalho, foi densenvolvido esse teste para o modelo de regressão beta com dispersão fixa e variável, como também, foram propostos um melhoramento nesse teste baseados na metodologia bootstrap e um novo teste, também com base em influência local, mas considerando outro esquema de perturbação, a perturbação no parâmetro de precisão no modelo de regressão beta com dispersão fixa. O desempenho desses testes foram avaliados com base no tamanho e poder. Por fim, aplicamos a teoria desenvolvida a um conjunto de dados reais. / The class of beta regression models introduced by Ferrari & Cribari-Neto (2004) is very useful for modelling rates and proportions. The proposed model by the authors is based on the assumption that the response variable is beta distributed with indexed by mean and dispersion parameters. After fitting a regression model is very important to carry out the diagnostic analysis in sense that, verifying possible deviations of the model assumptions, as well as detect possible observations that cause disproportionate influence on the parameter estimates. The local influence analysis introduced by Cook (1986) is an approach that objective assess the influence of observations. Based on local influence method, Zhu & Zhang (2004) proposed a hypothesis test to detect the degree of discrepancy between the supposed model and the underlying model from which the data is generated. In this work, was developed this test for the beta regression model with fixed and varying dispersion, as well as, we proposed in addition, an improvement of this test based on bootstrap methodology and a new test, also based on local influence, but considering other perturbation scheme, the perturbation of the precision parameter in beta regression model with fixed dispersion. The performance of these tests were evaluated based on size and power. Finally, we applied the theory developed to a set of real data.

Diploma thesis abstract Thesis title: Statistical inference in multivariate distributions based on copula models Author: Vojtěch Kika This diploma thesis aims for statistical inference in copula based models. Ba- sics of copula theory are described, followed by methods for statistical inference. These are divided into three main groups. First of them are parametric methods for copula parameter estimation which assume fully parametric structure, thus for both joint and marginal distributions. The second group consists of semi- parametric methods for copula parameter estimation which, unlike parametric methods, do not require parametric structure for marginal distributions. The last group describes goodness-of-fit tests used for testing the hypothesis that consi- dered copula belongs to some specific copula family. The thesis is accompanied by a simulation study that investigates the dependence of the observed coverage of the asymptotic confidence intervals for copula parameter on the sample size. Pseudolikelihood method was chosen for the simulation study since it is one of the most popular semiparametric methods. It is shown that sample size of 50 seems to be sufficient for the observed coverage to be close to the theoretical one. For Frank and Gumbel-Hougaard copula families even sample size of 30 gives us...

Une classe générale de statistiques de rangs basées sur la fonction caractéristique est introduite afin de tester l'hypothèse composite d'appartenance à une famille de copules multidimensionnelles. Ces statistiques d'adéquation sont définies comme des distances fonctionnelles de type L_2 pondérées entre une version non paramétrique et une version semi-paramétrique de la fonction caractéristique que l'on peut associer à une copule. Il est démontré que ces statistiques de test se comportent asymptotiquement comme des V-statistiques dégénérées d'ordre quatre et que leurs lois limites s'expriment en termes de sommes pondérées de variables khi-deux indépendantes. La convergence des tests sous des alternatives générales est établie, de même que la validité du bootstrap paramétrique pour le calcul de valeurs critiques. Le comportement des nouveaux tests sous des tailles d'échantillons faibles et modérées est étudié à l'aide de simulations et est comparé à celui d'un test concurrent fondé sur la copule empirique. La méthodologie est finalement illustrée sur un jeu de données à plusieurs dimensions.

Master of Science / Department of Statistics / Shie-Shien Yang / In fitting linear models, R[superscript]2 statistic has been wildly used as one of the measures to assess the goodness-of-fit and prediction power of the model. Unlike fixed linear models, at this time there is no single universally accepted measure for assessing goodness-of-fit and prediction power of a linear mixed model. In this report, we reviewed seven different approaches proposed to define a measure analogous to the usual R[superscript]2 statistic for assessing mixed models. One of seven statistics,Rc, has both conditional and marginal versions. Association mapping is an efficient way to link the genotype data with the phenotype diversity. When applying the R[superscript]2 statistic to the association mapping application, it can determine how well genetic polymorphisms, which are the explanatory variables in the mixed models, explain the phenotypic variation, which is the dependent variation. A linear mixed model method recently has been developed to control the spurious associations due to population structure and relative kinship among individuals of an association mapping. We assess seven definitions of R[superscript]2 statistic for the linear mixed model using data from two empirical association mapping samples: a sample with 277 diverse maize inbred lines and a global sample of 95 Arabidopsis thaliana accessions using the new method. R[superscript]2[subscript]LR statistic derived from the log-likelihood principle follows all the criterions of R[superscript]2 statistic and can be used to understand the overlap between population structure and relative kinship in controlling for sample relatedness. From our results,R[superscript]2[subscript]LR statistic is an appropriate R[superscript]2 statistic for comparing models with different fixed and random variables. Therefore, we recommend using RLR statistic for linear mixed models in association mapping.

There are some existing commonly used goodness-of-fit tests, such as the Kolmogorov-Smirnov test, the Cramer-Von Mises test, and the Anderson-Darling test. In addition, a new goodness-of-fit test named G test was proposed by Chen and Ye (2009). The purpose of this thesis is to compare the performance of some goodness-of-fit tests by comparing their power. A goodness-of-fit test is usually used when judging whether or not the underlying population distribution differs from a specific distribution. This research focus on testing whether the underlying population distribution is an exponential distribution. To conduct statistical simulation, SAS/IML is used in this research. Some alternative distributions such as the triangle distribution, V-shaped triangle distribution are used. By applying Monte Carlo simulation, it can be concluded that the performance of the Kolmogorov-Smirnov test is better than the G test in many cases, while the G test performs well in some cases.

When an Ornstein-Uhlenbeck (or CAR(1)) process is observed at discrete times 0, h, 2h,··· [T/h]h, the unobserved driving process can be approximated from the ob- served process. Approximated increments of the driving process are used to test the assumption that the process is L\'evy-driven. Asymptotic behavior of the test statis- tic at high sampling frequencies is developed assuming that the model parameters are known. The behavior of the test statistics using an estimated parameter is also studied. If it can be concluded that the driving process is L\'evy, the empirical process of the approximated increments can then be used to carry out more precise tests of goodness-of-fit. For example, one can test whether the driving process can be modeled as a Brownian motion or a gamma process. In each case, performance of the proposed test is illustrated through simulation.