Modelos multiníveis : gaussiano e multinomial

• Ano de defesa: 2017
• Autor(a) principal: Suguiura, Tiago Peres da Silva
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Maringá; Brasil; Departamento de Estatística; Programa de Pós-Graduação em Bioestatística; UEM; Maringá, PR; Centro de Ciências Exatas
• Palavras-chave em Português: Modelos multiníveis; Modelagem estatística; Modelo Linear misto; Modelo linear generalizado misto; Modelo multinominal; Afastamento gengival vertical; Câncer de mama; Brasil.; Multilevel models; Multinomial models; Mixed model; Generalized mixed model; Vertical gingival retraction; Breast cancer; Brazil.; Ciências Exatas e da Terra; Estatística
• Link de acesso: http://repositorio.uem.br:8080/jspui/handle/1/4359

The use of multilevel models in data sets from dentistry and public health has grown recently due to the experiment characteristics. In dentistry, for example, "side"/"root" of a tooth nested within a tooth and the tooth nested within each patient. Considering the public health, with diseases, with rates that can be influenced by the geographic space and culture habits, and for this reason, statistical analysis assuming independence of observed units is inappropriate, therefore, methodologies which take in account this characteristic must be adopted, which is the case of multilevel modeling. This work was separated in two parts, first, to evaluate the vertical gingival retraction, with a comparison between models with and without hierarchical structure, and, in the second part, making a literature review on multinomial multilevel model theory with the goal of analyzing an ongoing research about the distribution of visits due to breast cancer in hospitals from SUS in the state of Paraná from 2008 to Oct/2016, considering the 8 sub-topographies of breast cancer according to the International Classification of Diseases. The results from the first study showed that the analyzed substance is not effective to increase the vertical gingival retraction, and the multilevel model had a better adjustment to the hierarchical structure of the experiment. For the second study a better fit of the multinomial multilevel model was verified to analyze the occurrences of breast cancer by ICD, and presented differences between the occurrences in terms of Health Regions (Macro Regions), Race and Age Range.