| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.43 MB | Adobe PDF |
Orientador(es)
Resumo(s)
O meu projeto tem como objetivo principal analisar os fatores associados à obesidade na população adulta portuguesa e desenvolver um modelo de previsão para identificar a presença de
obesidade. Utilizei dados obtidos num amplo inquérito realizado em 2015-2016 acerca dos hábitos alimentares e da atividade física dos portugueses, que também inclui informações demográficas e socioeconómicas da população portuguesa.
Comecei por selecionar as variáveis com possível interesse para o meu estudo e realizei uma
análise exploratória das mesmas. Apresento estatísticas descritivas e gráficos para cada variável, contribuindo assim para uma compreensão abrangente do conjunto de dados.
O meu modelo de previsão foi desenvolvido através do uso de Regressão Logística. Inicialmente, realizei testes qui-quadrado, bem como análises de correlações, para avaliar associações
entre a presença de obesidade e as variáveis independentes. Isto permitiu-me escolher as variáveis que, à partida, entrariam no meu modelo como variáveis explicativas. Após a implementação do modelo, procedi à interpretação do mesmo. Interpretei os coeficientes do modelo logístico e calculei odds-ratios para compreender a relação entre as variáveis independentes e a
presença de obesidade.
Por fim, efetuei uma avaliação do desempenho do meu modelo, incluindo a análise da curva
ROC (Receiver Operating Characteristic) para determinar a sensibilidade e a especificidade do
modelo para diferentes pontos de corte. Calculei a área sob a curva ROC como medida de desempenho geral do modelo 0,701, o que é um valor com alguma razoabilidade.
A escolha do ponto de corte pode ser ajustada com base nas necessidades específicas das políticas de educação e da intervenção em obesidade. Obviamente que o ponto de corte ótimo dependerá da prevalência da doença no futuro e o mesmo poderá ser ajustado consoante quisermos
aumentar a sensibilidade ou a especificidade do teste.
Em resumo, o meu projeto fornece uma análise abrangente dos fatores associados à obesidade
em uma população adulta e desenvolve um modelo de previsão que pode ser adaptado para
apoiar políticas de educação e intervenção em obesidade, considerando a sensibilidade, a especificidade e as implicações clínicas.
My project aims to analyze the factors associated with obesity in the Portuguese adult population and develop a prediction model to identify the presence of obesity. I used data obtained from a comprehensive survey conducted in 2015-2016 on the dietary habits and physical activity of the Portuguese, which also includes demographic and socioeconomic information of the Portuguese population. I started by selecting variables of possible interest for my study and conducted an exploratory analysis of them. I presented descriptive statistics and graphs for each variable, contributing to a comprehensive understanding of the dataset. My prediction model was developed using Logistic Regression. Initially, I performed chisquare tests as well as correlation analyses to assess associations between the presence of obesity and independent variables. This allowed me to choose the variables that would initially enter my model as explanatory variables. After implementing the model, I proceeded to interpret it. I interpreted the coefficients of the logistic model and calculated odds ratios to understand the relationship between independent variables and the presence of obesity. Finally, I conducted an evaluation of the performance of my model, including the analysis of the ROC (Receiver Operating Characteristic) curve to determine the sensitivity and specificity of the model for different cutoff points. I calculated the area under the ROC curve as a measure of the overall performance of the model 0.701, which is a reasonably good value. The choice of the cutoff point can be adjusted based on the specific needs of education and obesity intervention policies. Obviously, the optimal cutoff point will depend on the prevalence of the disease in the future, and it can be adjusted depending on whether we want to increase the sensitivity or specificity of the test. In summary, my project provides a comprehensive analysis of the factors associated with obesity in an adult population and develops a prediction model that can be adapted to support education and obesity intervention policies, considering sensitivity, specificity, and clinical implications.
My project aims to analyze the factors associated with obesity in the Portuguese adult population and develop a prediction model to identify the presence of obesity. I used data obtained from a comprehensive survey conducted in 2015-2016 on the dietary habits and physical activity of the Portuguese, which also includes demographic and socioeconomic information of the Portuguese population. I started by selecting variables of possible interest for my study and conducted an exploratory analysis of them. I presented descriptive statistics and graphs for each variable, contributing to a comprehensive understanding of the dataset. My prediction model was developed using Logistic Regression. Initially, I performed chisquare tests as well as correlation analyses to assess associations between the presence of obesity and independent variables. This allowed me to choose the variables that would initially enter my model as explanatory variables. After implementing the model, I proceeded to interpret it. I interpreted the coefficients of the logistic model and calculated odds ratios to understand the relationship between independent variables and the presence of obesity. Finally, I conducted an evaluation of the performance of my model, including the analysis of the ROC (Receiver Operating Characteristic) curve to determine the sensitivity and specificity of the model for different cutoff points. I calculated the area under the ROC curve as a measure of the overall performance of the model 0.701, which is a reasonably good value. The choice of the cutoff point can be adjusted based on the specific needs of education and obesity intervention policies. Obviously, the optimal cutoff point will depend on the prevalence of the disease in the future, and it can be adjusted depending on whether we want to increase the sensitivity or specificity of the test. In summary, my project provides a comprehensive analysis of the factors associated with obesity in an adult population and develops a prediction model that can be adapted to support education and obesity intervention policies, considering sensitivity, specificity, and clinical implications.
Descrição
Trabalho de Projeto de Mestrado, Matemática Aplicada à Economia e Gestão, 2024, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Obesidade Teste do qui-quadrado Regressão logística Atividade física Teses de mestrado - 2024
