The problem of testing multiple hypotheses at once is called the problem of multiple testing. We focused on comparing more than two means in one- way analysis of variance, also known as ANOVA. We dealt with the Tukey me- thod, the Hothorn-Bretz-Westfall method, the bootstrap-based methods and also the Bonferroni method and its modification by the Holm method, the last two methods being popular mainly for their simplicity. We focused in detail on the asymptotic behavior of these methods and then compared them using si- mulations in terms of compliance with the prescribed level and in terms of average strength. Bonferroni's method, which is conservative, is known to lose strength compared to other methods. However, its modification of Holm's method, which is also conservative, in some cases by its strength equates to other more complex methods. 1

The intent of this thesis is to explore and develop the quantification of AOS features, particularly the deficits of prosodic elements, lexical stress and duration. This study investigated whether PVI can be used as a sensitive tool for the differential diagnosis of AOS. Specifically, we sought to determine whether analysis of vowel length of stressed and unstressed syllables is helpful in differentiating between individuals with AOS and aphasia versus aphasia alone. Significant differences support the hypothesis that PVI, analyzed from vowel length, is uniquely affected in AOS. This yields reason for further research in prosodic deficits in differential diagnosis, as well as application of this theory for a speech-language pathologist.

<p>Most existing classification algorithms either consider all features as equally important (equal weights), or do not analyze consistency of weights assigned to features. When features are not equally important, assigning consistent weights is a not obvious task. In general we have two cases. The first case assumes that a given sample of data does not contain any clue about the importance of features, so the weights are provided by a pool of experts and they are usually inconsistent. The second case assumes that the given sample contains some information about features importance, hence we can derive the weights directly from the sample. In this thesis we deal with both cases. Pairwise Comparisons and Weighted Support Vector Machines are used for the first case. For the second case a new approach based on the observation that the feature importance could be determined by the discrimination power of features has been proposed. For the first case, we start with pairwise comparisons to rank the importance of features, then we use distance-based inconsistency reduction to refine the weights assessment and make comparisons more precise. As the next step we calculate the weights through the fully-consistent or almost consistent pairwise comparison tables. For the second case, a novel concept of feature domain overlappings has been introduced. It can measure the feature discrimination power. This model is based on the assumption that less overlapping means more discrimination ability, and produces weights characterizing the importance of particular features. For both cases Weighted Support Vector Machines are used to classify the data. Both methods have been tested using two benchmark data sets, Iris and Vertebal.</p> <p>The results were especially superior to those obtained without weights.</p> / Master of Computer Science (MCS)

Networks, generically, refer to any application of graph theory in economics. Consider an undirected graph where nodes represent players and links represent relationships between them. Players can both form and delete links by which we mean that they can both form new relationships and terminate existing ones. A stable network is one in which no incentives exist to change the network structure. There can be various forms of stability depending on how many links players are allowed to form or delete at a time. Under strong pairwise stability, each player is allowed to delete any number of links at a time while any pair of players can form one link at a time. We introduce a network-value function, which assigns to each possible network a certain value. The value is allocated according to the component-wise egalitarian allocation rule, which divides the value generated by a component equally among members of the component (where a component refers to a maximally connected subgraph). An efficient network is one that maximizes the network value function. We show that there is an underlying conflict between strong pairwise stability and efficiency. Efficient networks are not necessarily strongly pairwise stable. This conflict can be resolved only if value functions satisfy a certain property called "middlemen-security". We further find that there is a broad class of networks called "middlemen-free networks" for which the above condition is automatically satisfied under all possible value functions. We also look at three network applications. A peering contract is an arrangement between Internet Service Providers under which they exchange traffic with one another free of cost. We analyze incentives for peering contracts among Internet service providers using the notion of pairwise stability. A hierarchy is a directed graph with an explicit top-down structure where each pair of linked agents have a superior-subordinate relationship with each other. We apply the notion of conjunctive permission value to demonstrate the formation of hierarchical firms in a competitive labor market. Comparative or targeted advertising is defined as any form of advertising where a firm directly or indirectly names a competitor. We also examine a model of targeted advertising between oligopolistic firms using non-cooperative game theoretic tools. / Ph. D.

In an effort to produce new derivatives of cellulose for drug delivery applications, methods were developed to regioselectively modify C-6 halo cellulose esters to produce cationic derivatives via nucleophilic substitution. Reaction of C-6 substituted bromo and iodo cellulose with trialkylated amines and phosphines produced new cationic ammonium and phosphonium cellulose derivatives which can be explored as delivery agents for nucleic acids, proteins and other anionic drug molecules. It was anticipated that these new derivatives would not only be capable of complexing anionic drug molecules but would have greatly improved aqueous solubility compared to their precursors. The phosphonium derivatives described in this work are an obvious example of such improved solubility properties. Given the importance of cellulose derivatives in making amorphous dispersions with critical drugs, it has also been important to analyze commercially available polymers for the potential impact in oral drug delivery formulations. To do so pairwise blends of cellulosics and synthetic polymers commonly used as excipients were tested for miscibility using techniques such as DSC, mDSC, FTIR and film clarity. Miscible combinations highlight the potential to use combinations of polymers currently available commercially to provide drug delivery solutions for specific drug formulations. The use of melt extrusion in processing some of these drug/polymer dispersions provides a means of highlighting the capability for the use of these cellulosics in melt extruded amorphous dispersions. This solvent free, high pressure method significantly reduces cost and time and can be applied on a large scale. The analysis of long chain cellulose esters and ultimately the novel omega carboxy esters for melt processability significantly impacts the possibilities available for use of those excellent drug delivery agents on a much larger scale. / Ph. D.

Dans les systèmes d'aide à la décision, sont généralement à disposition des données numériques abondantes et éventuellement certaines connaissances contextuelles qualitatives, disponibles a priori ou fournies a posteriori par retour d'expérience. Les performances des approches de classification, en particulier spectrale, dépendent de l'intégration de ces connaissances dans leur conception. Les algorithmes de classification spectrale permettent de traiter la classification sous l'angle de coupes de graphe. Ils classent les données dans l'espace des vecteurs propres de la matrice Laplacienne du graphe. Cet espace est censé mieux révéler la présence de groupements naturels linéairement séparables. Dans ce travail, nous nous intéressons aux algorithmes intégrant des connaissances type contraintes de comparaison. L'espace spectral doit, dans ce cas, révéler la structuration en classes tout en respectant, autant que possible, les contraintes de comparaison. Nous présentons un état de l'art des approches spectrales semi-supervisées contraintes. Nous proposons un nouvel algorithme qui permet de générer un sous-espace de projection par optimisation d'un critère de multi-coupes normalisé avec ajustement des coefficients de pénalité dus aux contraintes. Les performances de l'algorithme sont mises en évidence sur différentes bases de données par comparaison à d'autres algorithmes de la littérature. Dans le cadre de la surveillance de l'écosystème marin, nous avons développé un système de classification automatique de cellules phytoplanctoniques, analysées par cytométrie en flux. Pour cela, nous avons proposé de mesurer les similarités entre cellules par comparaison élastique entre leurs signaux profils caractéristiques. / In the decision support systems, often, there a huge digital data and possibly some contextual knowledge available a priori or provided a posteriori by feedback. The performances of classification approaches, particularly spectral ones, depend on the integration of the domain knowledge in their design. Spectral classification algorithms address the problem of classification in terms of graph cuts. They classify the data in the eigenspace of the graph Laplacian matrix. The generated eigenspace may better reveal the presence of linearly separable data clusters. In this work, we are particularly interested in algorithms integrating pairwise constraints : constrained spectral clustering. The eigenspace may reveal the data structure while respecting the constraints. We present a state of the art approaches to constrained spectral clustering. We propose a new algorithm, which generates a subspace projection, by optimizing a criterion integrating both normalized multicut and penalties due to the constraints. The performances of the algorithms are demonstrated on different databases in comparison to other algorithms in the literature. As part of monitoring of the marine ecosystem, we developed a phytoplankton classification system, based on flow cytometric analysis. for this purpose, we proposed to characterize the phytoplanktonic cells by similarity measures using elastic comparison between their cytogram signals.

Objetivo. Comparar, teórica e empiricamente, dois estimadores do coeficiente de correlação intraclasse momento-produto de Pearson para pares repetidos Pi. O primeiro é o estimador \"natural\", obtido mediante a correlação momento-produto de Pearson para membros de uma mesma classe (rI) e o segundo, obtido como função de componentes de variância (icc). Métodos. Comparação teórica e empírica dos parâmetros e estimadores. A comparação teórica envolve duas definições do coeficiente de correlação intraclasse PI como medida de confiabilidade (*), para o caso de duas réplicas, assim como uma apresentação da técnica de análise de variância e a definição e interpretação dos estimadores ri e icc. A comparação empírica é realizada mediante um estudo de simulação Monte Carlo com a geração de pares de valores correlacionados segundo o coeficiente de correlação intraclasse, momento-produto de Pearson para pares repetidos. Os pares de valores são distribuídos segundo uma distribuição Normal bivariada, com valores do tamanho da amostra e da correlação intraclasse previamente fixados em: n= 15, 30 e 45 e pI = {O; 0,15; 0,30; 0,45; 0,60; 0,75; 0,9}. Resultados. Comparando-se o vício e o erro quadrático médio dos estimadores, bem como as amplitudes dos intervalos de confiança, tem-se como resultado que o vício de icc foi sempre menor que o vício de rI, mesmo ocorrendo com o erro quadrático médio. Conclusões. O icc é um estimador melhor, principalmente para n pequeno (por exemplo 15). Para valores maiores de n (30 ou mais), os estimadores produzem resultados iguais até a segunda casa decimal. / Objective. This thesis presents and compares, theoretically and empirically, two estimators of the intraclass correlation coefficient pI, defined as Pearson\'s pairwise intraclass correlation coefficient. The first is the \"natural\" estimator, obtained by Pearson\'s moment-product correlation for members of one class (rI) while the second was obtained as a function of components of variance (icc). Methods. Theoretical and empirical comparison of the parameters and estimators are performed. The theoretical comparison involves two definitions of the intrac1ass correlation coefficient pI as a measure of reliability (*) for two repeated measurements in the same class and the presentation of the technique of analysis of variance, as well as for the definition and interpretation of the estimators ri and icc. The empirical comparison was carried out by means of a Monte Carlo simulation study of pairs of correlated values according Pearson\'s pairwise correlation. The pairs of values follow a normal bivariate distribution, with correlation values and sample size previously fixed: n= 15, 30 e 45 and Pl = . Results. Bias and mean square error for the estimators were compared as well as the range of the intervals of confidence. The comparison shows that the bias of icc is always smaller than of rI This also applies to the mean square error. Conclusions. The icc is a better estimator, especially for n less than or equal to 15. For larger samples sízes (n 30 or more), the estimators produce results that are equal to the second decimal place. (*) Fórmula

Os bandos mistos de aves são agregações sociais complexas estáveis durante o tempo e espaço. Até hoje, a estrutura social dessas espécies foi descrita a partir de estudos subjetivos de campo ou a partir de compilações do número e intensidade das interações a nível de todo o grupo, sem considerar as interações par-a-par individualmente. Nossos objetivos foram buscar evidências de associações não-randômicas entre pares de espécies de aves, delimitar os grupos a partir das espécies com as associações mais fortes e verificar se há diferenças na estrutura social entre os habitat de floresta primária e secundária. Utilizamos dados de ocorrência das espécies coletados a partir de redes de neblina e gravação de vocalizações para identificar pares de espécies que foram co-detectadas mais frequentemente do que o esperado a partir do modelo nulo e compararamos a força dessa interação entre as florestas tropicais primária e secundária Amazônicas. Nós também utilizamos as associações par-a-par para construir as redes de interação social e suas mudanças entre os tipos de habitat. Nós encontramos muitas interações positivas fortes entre as espécies, mas nenhuma evidência de repulsão. As análises das redes de interação revelaram vários grupos de espécies que corroboram com grupos ecológios descritos na literatura. Além disso, tanto a estrutura da rede de interação como a força da interação se alteraram drasticamente com a perturbação do habitat, com formação de algumas associações novas, mas uma tendência geral para quebra de associações entre as espécies. Nossos resultados mostram que as interações sociais entre essas aves podem ser fortemente afetados pela degradação do habitat, sugerindo que a estabilidade das interações desenvolvida entre espécies é ameaçada pelos distúrbios causados pelo homem. / Inter-specific associations of birds are complex social phenomena, frequently detected and often stable over time and space. So far, the social structure of these associations has been largely deduced from subjective assessments in the field or by counting the number of inter-specific encounters at the whole-group level, without considering changes to individual pairwise interactions. Here, we look for evidence of non-random association between pairs of bird species, delimit groups of more strongly associated species and examine differences in social structure between old growth and secondary forest habitat. We used records of bird species detection from mist-netting capture and from acoustic recordings to identify pairwise associations that were detected more frequently than expected under a null distribution, and compared the strength of these associations between old-growth and secondary forest Amazonian tropical forest. We also used the pairwise strength associations to visualize the social network structure and its changes between habitat types. We found many strongly positive interactions between species, but no evidence of repulsion. Network analyses revealed several modules of species that broadly agree with the subjective groupings described in the ornithological literature. Furthermore, both network structure and association strength changed drastically with habitat disturbance, with the formation of a few new associations but a general trend towards the breaking of associations between species. Our results show that social grouping in birds is real and may be strongly affected by habitat degradation, suggesting that the stability of the associations is threatened by anthropogenic disturbance.

No description available.

Mobile sinks are vital in many wireless sensor applications for efficient data collection, data querying, and localized sensor reprogramming. Mobile sinks prolong the lifetime of a sensor network. However, when sensor networks with mobile sinks are deployed in a hostile environment, security became a critical issue. They become exposed to varieties of malicious attacks. Thus, anti threats schemes and security services, such as mobile sink?s authentication and pairwise key establishment, are essential components for the secure operation of such networks. Due to the sensors, limited resources designing efficient security schemes with low communication overhead to secure communication links between sensors and MS (Mobile Sink) is not a trivial task. In addition to the sensors limited resources, sink mobility required frequent exchange of cryptography information between the sensors and MS each time the MS updates its location which imposes extra communication overhead on the sensors. In this dissertation, we consider a number of security schemes for WSN (wireless sensor network) with MS. The schemes offer high network?s resiliency and low communication overhead against nodes capture, MS replication and wormhole attacks. We propose two schemes based on the polynomial pool scheme for tolerating nodes capture: the probabilistic generation key pre-distribution scheme combined with polynomial pool scheme, and the Q-composite generation key scheme combined with polynomial pool scheme. The schemes ensure low communication overhead and high resiliency. For anti MS replication attack scheme, we propose the multiple polynomial pools scheme that provide much higher resiliency to MS replication attack as compared to the single polynomial pool approach. Furthermore, to improve the network resiliency against wormhole attack, two defensive mechanisms were developed according to the MS mobility type. In the first technique, MS uses controlled mobility. We investigate the problem of using a single authentication code by sensors network to verify the source of MS beacons, and then we develop a defensive approach that divide the sensor network into different authentication code?s grids. In the second technique, random mobility is used by MS. We explore the use of different communication channels available in the sensor hardware combined with polynomial pool scheme.