Yick Doi Pei. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2001. / Includes bibliographical references (leaves 63-66). / Abstracts in English and Chinese. / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Raw score approach --- p.4 / Chapter 1.2 --- Residual approach --- p.4 / Chapter 1.3 --- Indicator approach --- p.5 / Chapter 1.4 --- Overview --- p.5 / Chapter 2 --- Application --- p.7 / Chapter 2.1 --- Data --- p.7 / Chapter 3 --- Modeling --- p.10 / Chapter 3.1 --- Linear Regression at Individual Level --- p.13 / Chapter 3.2 --- Linear Regression at Group Level --- p.21 / Chapter 3.3 --- Polytomous Logistic Model --- p.28 / Chapter 3.4 --- Mixed Effect Model --- p.35 / Chapter 3.5 --- Discrete Response Multilevel Model --- p.41 / Chapter 4 --- Conclusion --- p.51 / Appendix --- p.55 / Reference --- p.63

In this dissertation a study will be made of university performance in the Science Faculty of the University of Natal, Durban. In particular, we will develop models that can be used to predict the success rate of a student based on his or her matriculation results. These models will prove useful for selecting students to universities. They may also be used to assist sponsors, bursars and donors in allocating funds to deserving students. In addition, these models may be used to identify students who might experience difficulties in their studies at university. / Thesis (M.Sc.)-University of Natal, Durban, 1994.

Outcomes Based Education (OBE), has been, since its inception, fraught with problems. OBE in its very nature is complex. To fully embrace this method and ensure its success, schools must be able to make the necessary paradigm shift. This can only be achieved when schools receive relevant and empowering training, support and development. In other words, organisational development must be the key words. The aim of this study is to explore the impact of organisational development interventions on the implementation of OBE. The case study method was employed where it was realised that schools that received organisational development interventions, together with Outcomes Based Education, were able to implement this method with greater understanding, skill, and confidence. The investigation recommends an organisational development design that could be used instead of the cascade model, and provides suggestions on what can be done to ensure a more successful implementation process. / Educational Studies / M. Ed (Education Management)

Orientadores: Filidor E. Vilca Labra, Victor Hugo Lachos Davila / Tese (doutorado) - Universidade Estadual de Campinas, Instituto de Matematica, Estatistica e Computação Cientifica / Made available in DSpace on 2018-08-14T22:06:24Z (GMT). No. of bitstreams: 1 Zeller_CamilaBorelli_D.pdf: 2738820 bytes, checksum: d40d3df77a4b5d44de0f48a8f8afed01 (MD5) Previous issue date: 2009 / Resumo: Neste trabalho, estudamos alguns aspectos de estimação e diagnóstico de influência local (Cook, 1986) em modelos lineares, especificamente no modelo de regressão linear, no modelo linear misto e no modelo de Grubbs sob a classe de distribuições assimétricas misturas de escala skew-normal (SMSN) (Branco & Dey, 2001). Esta família de distribuições tem como membros particulares as versões simétrica e assimétrica das distribuições t-Student, slash e normal contaminada, todas com caudas mais pesadas que a distribuição normal, A estimação dos parâmetros será via o algoritmo EM (Dempster et al, 1977) e a análise de diagnóstico será baseada na técnica de dados aumentados que usa a esperança condicional da função log-verossimilhança dos dados aumentados (função-Q) proveniente do algoritmo EM, como proposta por Zhu & Lee (2001) e Lee & Xu (2004). Assim, pretendemos contribuir positivamente para desenvolvimento da área dos modelos lineares, estendendo alguns resultados encontrados na literatura, por exemplo, Pinheiro et al (2001), Arellano-Valle et aí (2005), Osório (2006), Montenegro et al (2009a), Montenegro et al (2009b), Osório et al (2009), Lachos et aí (2010), entre outros. / Abstract: In this work, we study some aspects of the estimation and the diagnostics based on the local influence (Cook, 1986) in linear models under the class of scale mixtures of the skew-normal (SMSN) distribution, as proposed by Branco & Dey (2001). Specifically, we consider the linear regression model, the linear mixed model and the Grubbs' measurement error model. The SMSN class of distributions provides a useful generalization of the normal and the skew-normal distributions since it covers both the asymmetric and heavy-tailed distributions such as the skew-t, the skew-slash, the skew-contaminated normal, among others. The local influence analysis will be based on the conditional expectation of the complete-data log-likelihood function (function-Q) from the EM algorithm (Dempster et al, 1977) ), as proposed by Zhu & Lee (2001) and Lee & Xu (2004). We believe that the results of our work have contributed positively to the development of this area of linear models, since we have extended some results from the works of Pinheiro et al. (2001), Arellano-Valle et al. (2005), Osorio (2006), Montenegro et al. (2009a), Montenegro et al. (2009b), Osorio et al. (2009), Lachos et al. (2010), among others. / Doutorado / Método Estatístico / Doutor em Estatística

Orientador: Filidor Edilfonso Vilca Labra / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Matemática Estatística e Computação Científica / Made available in DSpace on 2018-08-25T19:08:47Z (GMT). No. of bitstreams: 1 HuairaContreras_CarlosAlberto_M.pdf: 2748699 bytes, checksum: fc8d02e2b19e638936faea1dec0b8ddc (MD5) Previous issue date: 2014 / Resumo: Modelos lineares são frequentemente usados em estatística para descrever a relação entre uma variável resposta e uma ou mais variáveis explicativas, onde geralmente os erros são assumidos como normalmente distribuídos. Além disso, em modelos de regressão linear assume-se que o mesmo modelo linear é válido para todo o conjunto de dados. O modelo pode mudar após um ponto específico e assim um modelo linear com um ponto de mudança poderá ser apropriado para o conjunto de dados. O principal objetivo deste trabalho é estudar alguns aspectos de estimação e análise de diagnóstico em modelos de regressão linear com ponto de mudança sob distribuições de mistura de escala normal. A análise de diagnóstico é baseada nos trabalhos de Cook (1986) e Zhu & Lee (2001). Os resultados obtidos representam uma extensão de alguns resultados apresentados na literatura, ver por exemplo Chen (1998) e Osorio & Galea (2005). Finalmente, estudos de simulação através de simulações Monte Carlo são realizados e exemplos numéricos são apresentados para ilustrar os resultados propostos / Abstract: Linear models are widely used in statistics to describe the relationship between a response variable and one or more explanatory variables, where usually it is assumed the errors are normally distributed. Moreover, in linear regression model is assumed that the same linear model holds for the whole data set, but this is not always valid. The model may change after a specific point, and so a linear model with a change point would be appropriate for data set. The main objective of work is to study some aspect of estimation and analysis of diagnostics in the regression linear with change point model under scale mixture of normal distributions. The analysis of diagnostics is based on the works of Cook (1986) and Zhu & Lee (2001). The results obtained represent a extension of some results obtained in the literature; see for example Chen (1998) and Osorio & Galea (2005). Finally, simulation studies are investigated through Monte Carlo simulations and numerical examples are presented to illustrate the proposed results / Mestrado / Estatistica / Mestre em Estatística

The role of inland fisheries as contributors to local and national economies in developing African countries is well documented. In South Africa, there is increasing interest in inland fisheries as vehicles for achieving national policy objectives including food security, livelihoods provision, poverty alleviation and economic development but there is surprisingly little literature on the history, current status, and potential of inland fishery resources. This lack of knowledge constrains the development of management strategies for ensuring the biological sustainability of these resources and the economic and social sustainability of the people that are dependent on them. In order to contribute to the knowledge base of inland fisheries in South Africa this thesis: (1) presents an exhaustive review of the available literature on inland fisheries in South Africa; (2) describes the organisation of recreational anglers (the primary users of the resource); (3) compiles recreational angling catch records and scientific gill net survey data, and assesses the applicability of these data for providing estimates of fish abundance (catch-per-unit effort [CPUE]); and finally, (4) determines the potential for models of fish abundance using morphometric, edaphic, and climatic factors. The literature review highlighted the data-poor nature of South African inland fisheries. In particular information on harvest rates was lacking. A lack of knowledge regarding different inland fishery sectors, governance systems, and potential user conflicts was also found. Recreational anglers were identified as the dominant user group and catch data from this sector were identified as potential sources of fish abundance and harvest information. Formal freshwater recreational angling in South Africa is a highly organised, multi-faceted activity which is based primarily on angling for non-native species, particularly common carp Cyprinus carpio and largemouth bass Micropterus salmoides. Bank anglers constituted the largest number of formal participants (5 309 anglers affiliated to formal angling organisations) followed by bass anglers (1 184 anglers affiliated to formal angling organisations). The highly structured nature of organised recreational angling and dominant utilisation of inland fisheries resources by this sector illustrated not only the vested interest of anglers in the management and development of inland fisheries but also the role that anglers may play in future decision-making and monitoring through the dissemination of catch data from organised angling events. Generalised linear models (GLMs) and generalised additive models (GAMs) were used to standardise CPUE estimates from bass- and bank angling catch records, which provided the most suitable data, and to determine environmental variables which most influenced capture probabilities and CPUE. Capture probabilities and CPUE for bass were influenced primarily by altitude and conductivity and multiple regression analysis revealed that predictive models incorporating altitude, conductivity, surface area and capacity explained significant (p<0.05) amounts of variability in CPUE (53%), probability of capture (49%) and probability of limit bag (74%). Bank angling CPUE was influenced by conductivity, surface area and rainfall although an insignificant (p>0.05) amount of variability (63%) was explained by a predictive model incorporating these variables as investigations were constrained by small sample sizes and aggregated catch information. Scientific survey data provided multi-species information and highlighted the high proportion of non-native fish species in Eastern Cape impoundments. Gillnet catches were influenced primarily by species composition and were less subject to fluctuations induced by environmental factors. Overall standardised gillnet CPUE was influenced by surface area, conductivity and age of impoundment. Although the model fit was not significant at the p<0.05 level, 23% of the variability in the data was explained by a predictive model incorporating these variables. The presence of species which could be effectively targeted by gillnets was hypothesised to represent the most important factor influencing catch rates. Investigation of factors influencing CPUE in impoundments dominated by Clarias gariepinus and native cyprinids indicated that warmer, younger impoundments and smaller, colder impoundments produced higher catches of C. gariepinus and native cyprinids respectively. A predictive model for C. gariepinus abundance explained a significant amount of variability (77%) in CPUE although the small sample size of impoundments suggests that predictions from this model may not be robust. CPUE of native cyprinids was influenced primarily by the presence of Labeo umbratus and constrained by small sample size of impoundments and the model did not adequately explain the variability in the data (r² = 0.31, p>0.05). These results indicate that angling catch- and scientific survey data can be useful in providing predictions of fish abundance that are biologically realistic. However, more data over a greater spatial scale would allow for more robust predictions of catch rates. This could be achieved through increased monitoring of existing resource users, the creation of a centralised database for catch records from angling competitions, and increased scientific surveys of South African impoundments conducted by a dedicated governmental function.

Indiana University-Purdue University Indianapolis (IUPUI) / Joint models of longitudinal and survival outcomes have been used with increasing frequency in clinical investigations. Correct specification of fixed and random effects, as well as their functional forms is essential for practical data analysis. However, no existing methods have been developed to meet this need in a joint model setting. In this dissertation, I describe a penalized likelihood-based method with adaptive least absolute shrinkage and selection operator (ALASSO) penalty functions for model selection. By reparameterizing variance components through a Cholesky decomposition, I introduce a penalty function of group shrinkage; the penalized likelihood is approximated by Gaussian quadrature and optimized by an EM algorithm. The functional forms of the independent effects are determined through a procedure for structural discovery. Specifically, I first construct the model by penalized cubic B-spline and then decompose the B-spline to linear and nonlinear elements by spectral decomposition. The decomposition represents the model in a mixed-effects model format, and I then use the mixed-effects variable selection method to perform structural discovery. Simulation studies show excellent performance. A clinical application is described to illustrate the use of the proposed methods, and the analytical results demonstrate the usefulness of the methods.