Fish behaviour analysis is a fundamental research area in marine ecology as it is helpful for detecting environmental changes by observing unusual fish patterns or new fish behaviours. The traditional way of analysing fish behaviour is by visual inspection using human observers, which is very time consuming and also limits the amount of data that can be processed. Therefore, there is a need for automatic algorithms to identify fish behaviours by using computer vision and machine learning techniques. The aim of this thesis is to help marine biologists with their work. We focus on behaviour understanding and analysis of detected and tracked fish with unusual behaviour detection approaches. Normal fish trajectories exhibit frequently observed behaviours while unusual trajectories are outliers or rare trajectories. This thesis proposes 3 approaches to detecting unusual trajectories: i) a filtering mechanism for normal fish trajectories, ii) an unusual fish trajectory classification method using clustered and labelled data and iii) an unusual fish trajectory classification approach using a clustering based hierarchical decomposition. The rule based trajectory filtering mechanism is proposed to remove normal fish trajectories which potentially helps to increase the accuracy of the unusual fish behaviour detection system. The aim is to reject normal fish trajectories as much as possible while not rejecting unusual fish trajectories. The results show that this method successfully filters out normal trajectories with a low false negative rate. This method is useful to assist building a ground truth data set from a very large fish trajectory repository, especially when the amount of normal fish trajectories greatly dominates the unusual fish trajectories. Moreover, it successfully distinguishes true fish trajectories from false fish trajectories which result from errors by the fish detection and tracking algorithms. A key contribution of this thesis is the proposed flat classifier, which uses an outlier detection method based on cluster cardinalities and a distance function to detect unusual fish trajectories. Clustered and labelled data are used to select feature sets which perform best on a training set. To describe fish trajectories 10 groups of trajectory descriptions are proposed which were not previously used for fish behaviour analysis. The proposed flat classifier improved the performance of unusual fish detection compared to the filtering approach. The performance of the flat classifier is further improved by integrating it into a hierarchical decomposition. This hierarchical decomposition method selects more specific features for different trajectory clusters which is useful considering the trajectory variety. Significantly improved results were obtained using this hierarchical decomposition in comparison to the flat classifier. This hierarchical framework is also applied to classification of more general imbalanced data sets which is a key current topic in machine learning. The experiments showed that the proposed hierarchical decomposition method is significantly better than the state of art classification methods, other outlier detection methods and unusual trajectory detection methods. Furthermore, it is successful at classifying imbalanced data sets even though the majority and minority classes contain varieties, and classes overlap which is frequently seen in real-world applications. Finally, we explored the benefits of active learning in the context of the hierarchical decomposition method, where active learning query strategies choose the most informative training data. A substantial performance gain is possible by using less labelled training data compared to learning from larger labelled data sets. Additionally, active learning with feature selection is investigated. The results show that feature selection has a positive effect on the performance of active learning. However, we show that random selection can be as effective as popular active learning query strategies in combination with active learning and feature selection, especially for imbalanced set classification.

Background: Transferring cost-effectiveness information between geographic domains offers the potential for more efficient use of analytical resources. However, it is difficult for decision-makers to know when they can rely on costeffectiveness evidence produced for another context. Objectives: This thesis explores the transferability of economic evaluation results produced for one geographic area to another location of interest, and develops an approach to identify factors to predict when this is appropriate. Methods: Multilevel statistical models were developed for the integration of published international costeffectiveness data to assess the impact of contextual effects on country-level; whilst controlling for baseline characteristics within, and across, a set of economic evaluation studies. Explanatory variables were derived from a list of factors suggested in the literature as possible constraints on the transferability of costeffectiveness evidence. The approach was illustrated using published estimates of the cost-effectiveness of statins for the primary and secondary prevention of cardiovascular disease from 67 studies and related to 23 geographic domains, together with covariates on data, study and country-level. Results: The proportion of variation at the country-level observed depends on the appropriate multilevel model structure and never exceeds 15% for incremental effects and 21% for incremental cost. Key sources of variability are patient and disease characteristics, intervention cost and a number of methodological characteristics defined on the data-level. There were fewer significant covariates on the study and country-levels. Conclusions: Analysis suggests that variability in cost-effectiveness data is primarily due to differences between studies, not countries. Further, comparing different models suggests that data from multinational studies severely underestimates country-level variability. Additional research is needed to test the robustness of these conclusions on other sets of cost-effectiveness data, to further explore the appropriate set of covariates, and to foster the development of multilevel statistical modelling for economic evaluation data in health.

Doctor of Philosophy / Department of Human Nutrition / Edgar Chambers IV / Preference mapping is a method that provides product development directions for developers to see a whole picture of products, liking and relevant descriptors in a target market. Many statistical methods and commercial statistical software programs offering preference mapping analyses are available to researchers. Because of numerous available options, there are two questions addressed in this research that most scientists must answer before choosing a method of analysis: 1) are the different methods providing the same interpretation, co-ordinate values and object orientation; and 2) which method and program should be used with the data provided? This research used data from paint, milk and fragrance studies, representing complexity from lesser to higher. The techniques used are principal component analysis, multidimensional preference map (MDPREF), modified preference map (PREFMAP), canonical variate analysis, generalized procrustes analysis and partial least square regression utilizing statistical software programs of SAS, Unscrambler, Senstools and XLSTAT. Moreover, the homogeneousness of consumer data were investigated through hierarchical cluster analysis (McQuitty’s similarity analysis, median, single linkage, complete linkage, average linkage, and Ward’s method), partitional algorithm (k-means method), nonparametric method versus four manual clustering groups (strict, strict-liking-only, loose, loose-liking-only segments). The manual clusters were extracted according to the most frequently rated highest for best liked and least liked products on hedonic ratings. Furthermore, impacts of plotting preference maps for individual clusters were explored with and without the use of an overall mean liking vector. Results illustrated various statistical software programs were not similar in their oriented and co-ordinate values, even when using the same preference method. Also, if data were not highly homogenous, interpretation could be different. Most computer cluster analyses did not segment consumers relevant to their preferences and did not yield as homogenous clusters as manual clustering. The interpretation of preference maps created by the highest homogeneous clusters had little improvement when applied to complicated data. Researchers should look at key findings from univariate data in descriptive sensory studies to obtain accurate interpretations and suggestions from the maps, especially for external preference mapping. When researchers make recommendations based on an external map alone for complicated data, preference maps may be overused.

Les zéolites, aluminosilicates cristallisés microporeux, sont largement utilisés en raffinage, en pétrochimie et en conversion de la biomasse. En raison du faible diamètre des micropores, limitations diffusionnelles et effets de confinement peuvent favoriser la formation de sous-Produits non désirés. L'introduction de mésopores par désalumination et/ou désilication ("zéolites hiérarchisées") peut diminuer ces phénomènes. Cependant, les mécanismes ces réactions restent méconnus à l'échelle moléculaire. Par calculs quantiques périodiques, au niveau de la théorie de la fonctionnelle de la densité (DFT) et selon une approche hybride QM/QM, nous avons pu mettre en évidence l'importance de l'attaque de la molécule d'eau sur l'atome d'aluminium, qui se fait en anti par rapport au site acide de Brønsted. Des structures d'Al penta ou tetra coordinées ont aussi été suggérées expérimentalement comme précurseurs de la désalumination. Malgré une forte hétérogénéité structurale des sites T, l'élucidation des chemins réactionnels et les énergies d’activation des étapes d’hydrolyse des liaisons Al-O (70-100 kJ/mol) dans les systèmes zéolitiques investigués (MOR, FAU, MFI, CHA) nous a permis d'établir des corrélations du type Brønsted-Evans-Polanyi. Ces corrélations permettent d'estimer et prédire des énergies d'activation par le biais de la thermodynamique, donne ainsi une prédiction aisée des sites T sensibles à la désalumination. Un autre facteur clé pour la compréhension de la désalumination est l'effet de confinement sur l'espèce aluminique extra-Réseau générée (EFAL), exercé par les cavités. Finalement nous avons pu montrer que le chemin réactionnel de désalumination et désilication, consécutif ou simultané, , est thermodynamiquement plus favorable qu'une simple désalumination ce qui est en accord avec les propositions mécanistiques de la littérature sur la genèse de mésopores par démétallation. / Zeolites are crystalline microporous aluminosilicates widely used in refining, petrochemistry and biomass conversion. However, diffusion limitation and confinement effect can promote the formation of undesired products. The introduction of mesopores by dealumination and/or desilication ("hierarchical zeolites") is a possible solution widely used experimentally. Nevertheless, the mechanisms of these demetallation reactions are poorly described at the molecular scale. We determine the mechanisms of the formation of extraframework Al species (EFAL) for zeotypes MOR, FAU, MFI and CHA occurring during the dealumination process, possibly associated with desilication. First-Principles periodic density functional theory (DFT) and hybrid QM/QM calculations have been employed in order to analyze full reaction paths leading to extraframework species and to quantify the activation energies of the determining steps. It has been demonstrated that the initiation of an Al-O(H) bond break takes place via water adsorption on the Al atom in anti-Position to the Brønsted acid site, via a penta- or tetra-Coordinated Al species. Such species are shown to be at the initiation of the Al dislodgement from the zeolitic framework. Despite a strong structural heterogeneity of T sites, we determined Brønsted-Evans-Polanyi (BEP) relationships for the entire dealumination pathway. Moreover, it is shown that not only the initiation and propagation mechanisms are primordial for the understanding of an Al extraction, but also the confinement effect on EFAL species within the zeolites cavities. Finally, from the energy profile of combined dealumination/desilication pathways, we show that it is thermodynamically favoured to extract extraframework Si species (EFSI) in the course of dealumination.

Philosophiae Doctor - PhD / The educational landscape in South Africa is unique and has also seen many changes since the dawn of democracy more than 20 years ago. The apartheid education system was marred by severe inequalities between schools and, for this reason, the democratic government post 1994 established a number of policies and interventions in an attempt to improve access, equity and quality between schools. The country has made significant advances in improving access to education. This is reflected in the Millennium Development Goals progress indicators showing that, as of 2013, almost all learners between the ages of 7 and 15 were enrolled in schools. While great strides have also been made with regard to equity, evidence shows that many schools in South Africa are still largely inequitable. Education quality, however, is an area that is still of grave concern and the matter requires much attention from educational stakeholders. International studies, such as the Trends in International Mathematics and Science Study (TIMSS) and the Progress in International Reading Literacy Study (PIRLS), use learner performance to measure the quality of the system. Such studies consistently report that South Africa is performing poorly and that large inequalities still exist between schools in the country. Improved quality is associated with effective schools and, in South Africa, only 20% of schools have been found to be functional or effective. Much of research focussed on school effectiveness, both nationally and internationally, however has been explained by factors in the school, including the appropriateness of curriculum content, infrastructure, resources in the school and teacher content knowledge. These factors have been found to be strongly correlated with effective schools.

Le résumé en français n'a pas été communiqué par l'auteur. / Le résumé en anglais n'a pas été communiqué par l'auteur.

Student Number: 9811822T - MSc Dissertation - School of Computational and Applied Mathematics - Faculty of Science / This dissertation reviews population dynamic type models of viral infection and introduces some new models to describe strain competition and the infected cell lifecycle. Laboratory data from a recent clinical trial, tracking drug resistant virus in patients given a short course of monotherapy is comprehensively analysed, paying particular attention to reproducibility. A Bayesian framework is introduced, which facilitates the inference of model parameters from the clinical data. It appears that the rapid emergence of resistance is a challenge to popular unstructured models of viral infection, and this challenge is partly addressed. In particular, it appears that minimal ordinary differential equations, with their implicit exponential lifetime (constant hazard) distributions in all compartments, lack the short transient timescales observed clinically. Directions for future work, both in terms of obtaining more informative data, and developing more systematic approaches to model building, are identified.

This study aimed to investigate the relationship between psychological capital and the work-related attitudes of job satisfaction and organisational commitment recognising the hierarchical nature of the data. This relationship was examined in light of a supportive organisational climate as defined by supervisor support which played the role of a moderator in this relationship. Data was gathered using a number of structured questionnaires which were distributed to employees via an online link. The Psychological Capital Questionnaire (Luthans, Youssef & Avolio, 2007), Organisational Commitment Questionnaire (Mowday, Steers & Porter, 1982), Warr, Cook and Wall’s (1979) measure of job satisfaction and Eisenberger’s (1986) adapted measure of supervisor support were administered. A total of 14 departments participated in the study and 50 employees completed the questionnaires. A Hierarchical Linear Model analysis (HLM) was used to analyse the data along with Pearson product moment correlations and a two-way ANOVA. Results indicated that psychological capital was related moderately and positively to job satisfaction but was not related to organisational commitment. Supervisor support was related to both job satisfaction and organisational commitment. Finally supervisor support moderated the relationship between psychological capital and job satisfaction but no interaction was found for the relationship between psychological capital and organisational commitment as moderated by supervisor support. This paper concludes with a discussion of the results, implications of the findings, limitations and directions for future research.

Como os processos que definem distribuições de espécies operam em diferentes escalas espaciais, abordagens multi-escalares são necessárias para que a variabilidade do sistema seja considerada no desenvolvimento de desenhos amostrais. Estudos sobre padrões espaciais de distribuição são necessários para servir de base para monitoramento e avaliações de impactos. Este trabalho avaliou a variação espacial de comunidades de infralitoral, em costões rochosos de três ilhas na Estação Ecológica dos Tupinambás. Os padrões espaciais foram investigados em quatro escalas, que variam de poucos metros a dezenas de quilômetros. Foram amostradas as profundidades de 1 a 5 e de 5 a 10 metros, no verão e inverno de 2013 e verão de 2014. O recobrimento percentual das unidades biológicas foi extraído a partir de foto-quadrados de 50x50 cm, aleatorizados, seguindo um desenho amostral aninhado. Os dominantes foram analisados individualmente com análise de variância univariada. Algas Calcárias Articuladas (ACA) foram o grupo dominante em todas as ilhas, períodos e profundidades, além de influenciar padrões multivariados, evidenciados pelo PCA. O recobrimento médio de ACA variou de 36% a 89.56%, considerando toda a amostragem. Outros dominantes variaram de acordo com o período e profundidade de coleta, dentre eles, Sargassum sp., Codium intertextum e Asparagopsis taxiformis. Todas as análises multivariadas (PERMANOVA, Pairwise Comparisons e nMDS) revelaram alta variação entre ilhas. Todas as outras escalas apresentaram variabilidade significativa, exceto a de poucos metros. Estimativas de variação revelaram que a variação residual e a entre ilhas foram sempre maiores do que as das outras escalas, confirmando a heterogeneidade intrínseca em pequena escala e a complexidade de comunidades insulares. O monitoramento nesta UC deverá incorporar todas as escalas investigadas. / As processes that define species distributions operate at different spatial scales, multi-scale approaches are needed to account for the system\'s natural variability, when developing sampling designs. Research about spatial patterns is essential to generate data to drive monitoring initiatives and impact assessments. The present study evaluated the spatial variability of subtidal rocky shore communities in three islands within the Ecological Station of Tupinambás. Spatial patterns were investigated across four scales, ranging from few meters to tens of kilometers, from 1 to 5 and from 5 to 10 m depth, in the summer and winter of 2013 and summer of 2014. Percent cover of the biological units was evaluated from photo-quadrats of 50x50 cm, randomly scattered, following a nested design. Dominant species and morphological groups were analyzed individually by univariate analyses of variance. Articulated Calcareous Algae (ACA) were the dominant group in all islands, depths and periods, and the main driver of multivariate patterns, as evidenced by the PCA. Mean cover of ACA varied from 36% to 89.56%, considering all occasions. Other dominant species were the algae Sargassum sp., Codium intertextum and Asparagopsis taxiformis, which varied according to sampling period and depth. All multivariate procedures (PERMANOVA, Pairwise Comparisons and nMDS plots) showed the highest variability among islands. All the other scales, except of few meters, presented significant variability. Estimates of variation showed that residual and intra-island variability were always higher than in other scales, confirming the intrinsic small-scale patchiness of marine assemblages and the complexity of insular communities. Monitoring efforts in this MPA should incorporate all the examined scales.

Os métodos de estimação para modelos de fragilidade vêm sendo bastante discutidos na literatura estatística devido a sua grande utilização em estudos de Análise de Sobrevivência. Vários métodos de estimação de parâmetros dos modelos foram desenvolvidos: procedimentos de estimação baseados no algoritmo EM, cadeias de Markov de Monte Carlo, processos de estimação usando verossimilhança parcial, verossimilhança penalizada, quasi-verossimilhança, entro outros. Uma alternativa que vem sendo utilizada atualmente é a utilização da verossimilhança hierárquica. O objetivo principal deste trabalho foi estudar as vantagens e desvantagens da verossimilhança hierárquica para a inferência em modelos de fragilidade em relação a verossimilhança penalizada, método atualmente mais utilizado. Nós aplicamos as duas metodologias a um banco de dados real, utilizando os pacotes estatísticos disponíveis no software R, e fizemos um estudo de simulação, visando comparar o viés e o erro quadrático médio das estimativas de cada abordagem. Pelos resultados encontrados, as duas metodologias apresentaram estimativas muito próximas, principalmente para os termos fixos. Do ponto de vista prático, a maior diferença encontrada foi o tempo de execução do algoritmo de estimação, muito maior na abordagem hierárquica. / Estimation procedures for frailty models have been widely discussed in the statistical literature due its widespread use in survival studies. Several estimation methods were developed: procedures based on the EM algorithm, Monte Carlo Markov chains, estimation processes based on parcial likelihood, penalized likelihood and quasi-likelihood etc. An alternative currently used is the hierarchical likelihood. The main objective of this work was to study the hierarchical likelihood advantages and disadvantages for inference in frailty models when compared with the penalized likelihood method, which is the most used one. We applied both approaches to a real data set, using R packages available. Besides, we performed a simulation study in order to compare the methods through out the bias and the mean square error of the estimators. Both methodologies presented very similar estimates, mainly for the fixed effects. In practice, the great difference was the computational cost, much higher in the hierarchical approach.