Master of Science / Department of Statistics / Paul I. Nelson / Quantiles are useful in describing distributions of component lifetimes. Data, consisting of the lifetimes of sample units, used to estimate quantiles are often censored. Right censoring, the setting investigated here, occurs, for example, when some test units may still be functioning when the experiment is terminated. This study investigated and compared the performance of parametric and nonparametric estimators of quantiles from right censored data generated from Weibull and Lognormal distributions, models which are commonly used in analyzing lifetime data. Parametric quantile estimators based on these assumed models were compared via simulation to each other and to quantile estimators obtained from the nonparametric Kaplan- Meier Estimator of the survival function. Various combinations of quantiles, censoring proportion, sample size, and distributions were considered. Our simulation show that the larger the sample size and the lower the censoring rate the better the performance of the estimates of the 5th percentile of Weibull data. The lognormal data are very sensitive to the censoring rate and we observed that for higher censoring rates the incorrect parametric estimates perform the best. If you do not know the underlying distribution of the data, it is risky to use parametric estimates of quantiles close to one. A limitation in using the nonparametric estimator of large quantiles is their instability when the censoring rate is high and the largest observations are censored. Key Words: Quantiles, Right Censoring, Kaplan-Meier estimator

Full weighted least squares (full WLS) and robust weighted least squares (robust WLS) are currently the two primary estimation methods designed for structural equation modeling with ordinal observed variables. These methods assume that continuous latent variables were coarsely categorized by the measurement process to yield the observed ordinal variables, and that the model proposed by the researcher pertains to these latent variables rather than to their ordinal manifestations. Previous research has strongly suggested that robust WLS is superior to full WLS when models are correctly specified. Given the realities of applied research, it was critical to examine these methods with misspecified models. This Monte Carlo simulation study examined the performance of full and robust WLS for two-factor, eight-indicator confirmatory factor analytic models that were either correctly specified, overspecified, or misspecified in one of two ways. Seven conditions of five-category indicator distribution shape at four sample sizes were simulated. These design factors were completely crossed for a total of 224 cells. Previously findings of the relative superiority of robust WLS with correctly specified models were replicated, and robust WLS was also found to perform better than full WLS given overspecification or misspecification. Robust WLS parameter estimates were usually more accurate for correct and overspecified models, especially at the smaller sample sizes. In the face of misspecification, full WLS better approximated the correct loading values whereas robust estimates better approximated the correct factor correlation. Robust WLS chi-square values discriminated between correct and misspecified models much better than full WLS values at the two smaller sample sizes. For all four model specifications, robust parameter estimates usually showed lower variability and robust standard errors usually showed lower bias. These findings suggest that robust WLS should likely remain the estimator of choice for applied researchers. Additionally, highly leptokurtic distributions should be avoided when possible. It should also be noted that robust WLS performance was arguably adequate at the sample size of 100 when the indicators were not highly leptokurtic. / text

L'objectif de cette thèse est d'étudier le comportement local d'une mesure de probabilité, notamment au travers d'un indice de régularité locale. Dans la première partie, nous établissons la normalité asymptotique de l'estimateur des kn plus proches voisins de la densité et de l'histogramme. Dans la deuxième, nous définissons un estimateur du mode sous des hypothèses affaiblies. Nous montrons que l'indice de régularité intervient dans ces deux problèmes. Enfin, nous construisons dans une troisième partie différents estimateurs pour l'indice de régularité à partir d'estimateurs de la fonction de répartition, dont nous réalisons une revue bibliographique.

Dans un contexte de test multiple, nous considérons un modèle de mélange semi-paramétrique avec deux composantes. Une composante est supposée connue et correspond à la distribution des p-valeurs sous hypothèse nulle avec probabilité a priori p. L'autre composante f est nonparamétrique et représente la distribution des p-valeurs sous l'hypothèse alternative. Le problème d'estimer les paramètres p et f du modèle apparaît dans les procédures de contrôle du taux de faux positifs (``false discovery rate'' ou FDR). Dans la première partie de cette dissertation, nous étudions l'estimation de la proportion p. Nous discutons de résultats d'efficacité asymptotique et établissons que deux cas différents arrivent suivant que f s'annule ou non surtout un intervalle non-vide. Dans le premier cas (annulation surtout un intervalle), nous présentons des estimateurs qui convergent \`{a} la vitesse paramétrique, calculons la variance asymptotique optimale et conjecturons qu'aucun estimateur n'est asymptotiquement efficace (i.e atteint la variance asymptotique optimale). Dans le deuxième cas, nous prouvons que le risque quadratique de n'importe quel estimateur ne converge pas à la vitesse paramétrique. Dans la deuxième partie de la dissertation, nous nous concentrons sur l'estimation de la composante inconnue nonparamétrique f dans le mélange, en comptant sur un estimateur préliminaire de p. Nous proposons et étudions les propriétés asymptotiques de deux estimateurs différents pour cette composante inconnue. Le premier estimateur est un estimateur à noyau avec poids aléatoires. Nous établissons une borne supérieure pour son risque quadratique ponctuel, en montrant une vitesse de convergence nonparamétrique classique sur une classe de Holder. Le deuxième estimateur est un estimateur du maximum de vraisemblance régularisée. Il est calculé par un algorithme itératif, pour lequel nous établissons une propriété de décroissance d'un critère. De plus, ces estimateurs sont utilisés dans une procédure de test multiple pour estimer le taux local de faux positifs (``local false discovery rate'' ou lfdr). / In a multiple testing context, we consider a semiparametric mixture model with two components. One component is assumed to be known and corresponds to the distribution of p-values under the null hypothesis with prior probability p. The other component f is nonparametric and stands for the distribution under the alternative hypothesis. The problem of estimating the parameters p and f of the model appears from the false discovery rate control procedures. In the first part of this dissertation, we study the estimation of the proportion p. We discuss asymptotic efficiency results and establish that two different cases occur whether f vanishes on a non-empty interval or not. In the first case, we exhibit estimators converging at parametric rate, compute the optimal asymptotic variance and conjecture that no estimator is asymptotically efficient (i.e. attains the optimal asymptotic variance). In the second case, we prove that the quadratic risk of any estimator does not converge at parametric rate. In the second part of the dissertation, we focus on the estimation of the nonparametric unknown component f in the mixture, relying on a preliminary estimator of p. We propose and study the asymptotic properties of two different estimators for this unknown component. The first estimator is a randomly weighted kernel estimator. We establish an upper bound for its pointwise quadratic risk, exhibiting the classical nonparametric rate of convergence over a class of Holder densities. The second estimator is a maximum smoothed likelihood estimator. It is computed through an iterative algorithm, for which we establish a descent property. In addition, these estimators are used in a multiple testing procedure in order to estimate the local false discovery rate.

This dissertation studies three classes of estimators for the asymptotic variance parameter of a stationary stochastic process. All estimators are based on the concept of data "re-use" and all transform the output process into functions of an approximate Brownian motion process. The first class of estimators consists folded standardized time series area and Cramér-von Mises (CvM) estimators. Detailed expressions are obtained for their expectation at folding levels 0 and 1; those expressions explain the puzzling increase in small-sample bias as the folding level increases. In addition, we use batching and linear combinations of estimators from different levels to produce estimators with significantly smaller variance. Finally, we obtain very accurate approximations of the limiting distributions of batched folded estimators. These approximations are used to compute confidence intervals for the mean and variance parameter of the underlying stochastic process. The second class --- folded overlapping area estimators --- are computed by averaging folded versions of the standardized time series corresponding to overlapping batches. We establish the limiting distributions of the proposed estimators as the sample size tends to infinity. We obtain statistical properties of these estimators such as bias and variance. Further, we find approximate confidence intervals for the mean and variance parameter of the process by approximating the theoretical distributions of the proposed estimators. In addition, we develop algorithms to compute these estimators with only order-of-sample-size work. The third class --- reflected area and CvM estimators --- are computed from reflections of the original sample path. We obtain the expected values and variance of individual estimators. We show that it is possible to obtain linear combinations of reflected estimators with smaller variance than the variance of each constituent estimator, often at no cost in bias. A quadratic optimization problem is solved to find an optimal linear combination of estimators that minimizes the variance of the linearly combined estimator. For all classes of estimators, we provide Monte Carlo examples to show that the estimators perform as well in practice as advertised by the theory.

Made available in DSpace on 2014-06-12T18:29:15Z (GMT). No. of bitstreams: 2 arquivo4279_1.pdf: 1161561 bytes, checksum: 80aee0b17f88de11dd7d0999ad1594a1 (MD5) license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2008 / Esta tese de doutorado trata da realização de inferências no modelo de regressão linear sob heteroscedasticidade de forma desconhecida. No primeiro capítulo, nós desenvolvemos estimadores intervalares que são robustos à presença de heteroscedasticidade. Esses estimadores são baseados em estimadores consistentes de matrizes de covariâncias propostos na literatura, bem como em esquemas bootstrap. A evidência numérica favorece o estimador intervalar HC4. O Capítulo 2 desenvolve uma seqüência corrigida por viés de estimadores de matrizes de covariâncias sob heteroscedasticidade de forma desconhecida a partir de estimador proposto por Qian eWang (2001). Nós mostramos que o estimador de Qian-Wang pode ser generalizado em uma classe mais ampla de estimadores consistentes para matrizes de covariâncias e que nossos resultados podem ser facilmente estendidos a esta classe de estimadores. Finalmente, no Capítulo 3 nós usamos métodos de integração numérica para calcular as distribuições nulas exatas de diferentes estatísticas de testes quasi-t, sob a suposição de que os erros são normalmente distribuídos. Os resultados favorecem o teste HC4

The purpose of this study is to determine the effect of three improvement methods on nonparametric kernel regression estimators. The improvement methods are applied to the Nadaraya-Watson estimator with crossvalidation bandwidth selection, the Nadaraya-Watson estimator with plug-in bandwidth selection, the local linear estimator with plug-in bandwidth selection and a bias corrected nonparametric estimator proposed by Yao (2012). The di erent resulting regression estimates are evaluated by minimising a global discrepancy measure, i.e. the mean integrated squared error (MISE). In the machine learning context various improvement methods, in terms of the precision and accuracy of an estimator, exist. The rst two improvement methods introduced in this study are bootstrapped based. Bagging is an acronym for bootstrap aggregating and was introduced by Breiman (1996a) from a machine learning viewpoint and by Swanepoel (1988, 1990) in a functional context. Bagging is primarily a variance reduction tool, i.e. bagging is implemented to reduce the variance of an estimator and in this way improve the precision of the estimation process. Bagging is performed by drawing repetitive bootstrap samples from the original sample and generating multiple versions of an estimator. These replicates of the estimator are then used to obtain an aggregated estimator. Bragging stands for bootstrap robust aggregating. A robust estimator is obtained by using the sample median over the B bootstrap estimates instead of the sample mean as in bagging. The third improvement method aims to reduce the bias component of the estimator and is referred to as boosting. Boosting is a general method for improving the accuracy of any given learning algorithm. The method starts of with a sensible estimator and improves iteratively, based on its performance on a training dataset. Results and conclusions verifying existing literature are provided, as well as new results for the new methods. / MSc (Statistics), North-West University, Potchefstroom Campus, 2015

The purpose of this study is to determine the effect of three improvement methods on nonparametric kernel regression estimators. The improvement methods are applied to the Nadaraya-Watson estimator with crossvalidation bandwidth selection, the Nadaraya-Watson estimator with plug-in bandwidth selection, the local linear estimator with plug-in bandwidth selection and a bias corrected nonparametric estimator proposed by Yao (2012). The di erent resulting regression estimates are evaluated by minimising a global discrepancy measure, i.e. the mean integrated squared error (MISE). In the machine learning context various improvement methods, in terms of the precision and accuracy of an estimator, exist. The rst two improvement methods introduced in this study are bootstrapped based. Bagging is an acronym for bootstrap aggregating and was introduced by Breiman (1996a) from a machine learning viewpoint and by Swanepoel (1988, 1990) in a functional context. Bagging is primarily a variance reduction tool, i.e. bagging is implemented to reduce the variance of an estimator and in this way improve the precision of the estimation process. Bagging is performed by drawing repetitive bootstrap samples from the original sample and generating multiple versions of an estimator. These replicates of the estimator are then used to obtain an aggregated estimator. Bragging stands for bootstrap robust aggregating. A robust estimator is obtained by using the sample median over the B bootstrap estimates instead of the sample mean as in bagging. The third improvement method aims to reduce the bias component of the estimator and is referred to as boosting. Boosting is a general method for improving the accuracy of any given learning algorithm. The method starts of with a sensible estimator and improves iteratively, based on its performance on a training dataset. Results and conclusions verifying existing literature are provided, as well as new results for the new methods. / MSc (Statistics), North-West University, Potchefstroom Campus, 2015

Approved for public release, distribution is unlimited / This thesis examines the magnitude of the deterrence effect associated with the implementation of the "zero tolerance" policy in the U.S. military in the early 1980s. The estimation of the deterrence effect is based on the estimation of linear probability models (LPM). A difference-in-difference estimator is obtained by comparing pre- and post-policy differences in drug use rates in the military and civilian sectors. The thesis uses data on drug participation drawn from the National Household Survey of Drug Abuse and the DoD Worldwide Health Survey. The study investigates the deterrence effect for the military as a whole, for each branch, for various age groups, and two different measures of drug participation. The results show that a significant deterrence effect appears to have been associated with the implementation of the "zero tolerance" and drug testing policy, especially for the past year drug participation rates. Additionally, there is evidence that individuals above 25 years old who are more educated and married have smaller drug participation rates than the rest of the population. / Lieutenant Commander, Hellenic Navy

In this thesis, we study re-weighting when estimating totals in survey sampling. The purpose of re-weighting is to adjust the structure of the sample in order to comply with the structure of the population (with respect to given auxiliary variables). We sum up some known results for methods of the traditional desin-based approach, more attention is given to the model-based approach. We generalize known asymptotic results in the model-based theory to a wider class of weighted estimators. Further, we propose a consistent estimator of asymptotic variance, which takes into consideration weights used in estimator of the total. This is in contrast to usually recommended variance estimators derived from the design-based approach. Moreover, the estimator is robust againts particular model misspecifications. In a simulation study, we investigate how the proposed estimator behaves in comparison with variance estimators which are usually recommended in the literature or used in practice. 1