Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
A finite mixture distribution for modelling overdispersion
Maria Iannario

Last modified: 2013-06-16


This paper describes a strategy for modelling overdispersion in survey data concerning marginal ranking variables. We assume that ordinal data have been generated by a mixture distribution, defined as CUBE model, where a specific parameter explains a possible presence of overdispersion. Asymptotic likelihood methods will be applied for efficient statistical inference. The method is primarily motivated and then implemented by a sample on the emergencies in a metropolitan area. The proposed model improves both fitting and interpretation for the social and urban problems involved.

Full Text: PDF