Modelos com variáveis latentes aplicados à mensuração de importância de atributos

• Ano de defesa: 2006
• Autor(a) principal: Samartini, André Luiz Silva
• Orientador(a): Bussab, Wilton de Oliveira
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Ponderação pela importância; Teoria de resposta ao item; Modelo de desdobramento graduado generalizado; Imputação de pesos
• Palavras-chave em Inglês: Item response theory; Importance weighting; Generalized graded unfolding model; Weighting imputation
• Link de acesso: https://hdl.handle.net/10438/2553

Many surveys use multi-items questionnaires to make an assessment of a service or product. In these surveys, each respondent, besides having to evaluate each item (attribute) according to the degree of satisfaction or agreement, must also evaluate its importance. It is possible, with these two informations, to create a global score of the product or service weighted by the importance of the items. The goal of this thesis is to model the importance through a Generalized Graded Unfolding Model, which is an Item Response Theory model. Results of a survey showed a good fit of the model, and simulations showed that it is possible to design experiments to reduce the number of items or response categories asked to the respondents without loss of information if this model is used.