Estudo comparativo de métodos de estimação do modelo de resposta gradual para dados de burnout em enfermeiras

• Ano de defesa: 2020
• Autor(a) principal: Maia, Juliana Marambaia
• Orientador(a): Novelli, Cibele Maria Russo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa Interinstitucional de Pós-Graduação em Estatística - PIPGEs
• Palavras-chave em Português: Teoria de resposta ao item; Modelo de resposta gradual; Inferência Bayesiana; Inferência clássica; Síndrome de burnout
• Palavras-chave em Inglês: Item response theory; Graded response model; Bayesian inference; Classical inference; Burnout syndrome
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/12848

The graded response model is an Item Response Theory (IRT) model where items admit a polytomous response widely known in the literature. The motivation for this study comes from a data set which belongs to the RN4CAST project related to the burnout syndrome. This psychological syndrome has a multidimensional configuration of emotional exhaustion, depersonalization and personal accomplishment. The burnout measurements are obtained from the Maslach Burnout Inventory, a 22-item questionnaire to be answered on a 7-point Likert scale. Our proposal in this work is a study of the unidimensional and multiunidimensional graded response model, under the frequentist and Bayesian approaches, motivated by the data from the burnout syndrome, and for that, a simulation study is carried out to verify the behavior of empirical form of the model. Models are fitted with the marginal maximum likelihood and the Monte Carlo Markov chain via Gibbs sampler methods. The simulation study shows that, in general, the Monte Carlo Markov chain via Gibbs sampler method produces good results in the estimation of the item parameters and latent trait of the unidimensional gradual response model. The results of the multiunidimensional gradual response model study yield good results as the sample size and test size increase.