Teses e dissertaçõeshttps://repositorio.ufscar.br/handle/ufscar/46002024-03-29T09:02:26Z2024-03-29T09:02:26ZUma abordagem estatística para a análise dos resultados das eleições presidenciaisLachos Olivares, Victor Eduardohttps://repositorio.ufscar.br/handle/ufscar/195472024-02-28T22:06:32Z2024-01-15T00:00:00ZUma abordagem estatística para a análise dos resultados das eleições presidenciais
Lachos Olivares, Victor Eduardo
Multiparty data has characteristics that make it compositional data such as a constant sum of components and a limited space known as simplex. Thus, the purpose of the work is to develop a methodology to analyze multi-party data from electoral elections considering their restricted nature. In this context, the proposed methodology consists of 8 steps: initially, we collect multi-party data and transform it into compositional data. Then, we apply the log-ratio transformation , removing the inherent constraints of compositional data.
Next, we employ principal component analysis (PCA) to reduce dimensionality and identify the principal components that retain most of the variation in the data. These components are analyzed based on two important metrics: loadings and scores. Given that the scores have different variability in the components, they are transformed between values of zero and one.
Subsequently, we propose the Beta regression model considering the scores as the response variable, and the human development indicators as the explanatory variables. The methodology is applied to multiparty data from the first round elections in Peru in 2021 and Brazil in 2022, allowing us to identify the main components and which covariates (health, education and income) are directly related to votes in different regions and states.
Finally, considering that data from presidential elections of Peru 2021 with two response variables, we propose a bivariate regression model via copulas and analyze the dependence structure between these variables.
2024-01-15T00:00:00ZModelagem de predição de crimes na região metropolitana de São PauloZhao, Wellington Yunahehttps://repositorio.ufscar.br/handle/ufscar/195302024-02-28T20:57:26Z2023-12-13T00:00:00ZModelagem de predição de crimes na região metropolitana de São Paulo
Zhao, Wellington Yunahe
The issue of public security is a challenge for Brazilian society, and crime is a major concern for the most populous state in the country, São Paulo. It is always desirable for the public administration to model and predict criminal trends, taking into account historical dates and the georeferencing of each municipality, meaning latitude and longitude. In this context, the use of spatiotemporal models to explain the relationship between predictors and crimes, as well as considering location, can be of great importance. One possible model is Spatial Autoregressive (SAR) modeling, which takes into account covariates and implicit spatial dependencies. In this work, based on the number of crimes, SAR modeling is used to describe and model the cities in the metropolitan region of São Paulo, Brazil, including the annual seasonality observed in the data. To visualize the data and develop modeling with the spatial neighborhood matrix, R packages such as spatialreg are used. The Lasso method is used to pre-select variables with greater significance, such as the number of inhabitants per household, the dropout rate, and the public elementary school dropout rate in the early years. Then, the SAR model is applied to include spatial information and enhance crime modeling. In general, this work focuses on developing spatiotemporal modeling for crimes in the state of São Paulo, identifying predictor variables that influence the quantity of crimes in a given municipality. In addition to the SAR model, artificial neural networks, such as multilayer models and Long Short-Term Memory (LSTM), are also used in the research, and compared with the SAR model. The goal of this dissertation is to develop predictive modeling considering spatiotemporal data for crimes in the metropolitan region of São Paulo, using predictor variables that influence the occurrence and quantity of crimes in a particular municipality. It is expected that the obtained results are useful for decision making by public administration, since the work creates a method to analyze crime patterns in a specific municipality, and also helps the city improve security issues through various social factors.
2023-12-13T00:00:00ZDiagnóstico e seleção de modelos com resposta binária e função de ligação assimétricaCoelho, Fabiano Rodrigueshttps://repositorio.ufscar.br/handle/ufscar/193712024-02-19T12:20:31Z2023-12-06T00:00:00ZDiagnóstico e seleção de modelos com resposta binária e função de ligação assimétrica
Coelho, Fabiano Rodrigues
For binary response variables, probit and logit link functions are widely used. However, when the data is imbalanced, traditional approaches may not be suitable. In this thesis, we consider the skew-probit link function as a potential alternative for models with binary response. The parameters are estimated through a Bayesian approach using Hamiltonian Monte Carlo, and residual analysis is developed. Additionally, an extension for the case of mixed models is presented, with parameter estimation performed through numerical integration. As a practical application, we analyze two datasets. In both applications, it is possible to observe, through model selection criteria, that the skew-probit regression model is more efficient than traditional approaches. Computationally, for the fixed-effects model, we use the Stan language adapted to the R software. In the mixed case, the INLA methodology is considered. Proposals for future research are also discussed.
2023-12-06T00:00:00ZModelagem via redes neurais de dados de sobrevivência de longa duração com dispersão não observadaTeh, Led Redhttps://repositorio.ufscar.br/handle/ufscar/190442023-12-20T11:55:15Z2023-12-08T00:00:00ZModelagem via redes neurais de dados de sobrevivência de longa duração com dispersão não observada
Teh, Led Red
Traditional models in survival analysis assume that every subject will eventually experience the event of interest in the study, such as death or disease recurrence, so the survival function is said to be proper. Cure rate model, which was first proposed seven decades ago, has since been used to account for the presence of cure fraction, this means that a certain fraction of the individuals will never experience the occurrence of an event of interest for which they can be treated as immune or cured subjects in the context of cancer treatment. In the literature, various cure rate models have been widely studied and commonly applied to structured data with small quantities of covariates. The use of convolutional neural network, a powerful deep learning technique for image processing problem, has become increasingly more common in the medical field in recent years. Medical images such as histological slides and magnetic resonance images (MRIs) are directly related to a patient's prognostic factors, therefore, it is reasonable to introduce these images as predictors in cure model. In this work, we extend upon the article of Xie and Yu (2021b) in which a neural network was used to model the unstructured predictor's effect in the promotion time cure model's setting to the cases of overdispersed data. We will call our extension as integrated negative binomial cure rate model, and its parameters will be estimated through the Expectation-Maximization algorithm.
2023-12-08T00:00:00Z