Tesis Doctorado
Similarity analysis in species sampling mixture models.
Autor
Navarrete, Carlos A
Institución
Resumen
Species SampLing Mixture Modeis (SSMMs) rely on inodeling the
data en top of a structure that considers the inherent clustering of the observations
under the hypothesis of prior exchangeability. This work proposes a
method to study the information given by the posterior clustering behaviour
of SSMMs, called Sirnilarity Analysis. It is based fundamentallv en decomposing
the similarity matrix obtained from a sample of the partitions in an
intrinsic and an extnnsic part, This gives valuable information about the individual
characteristics that explain the clustering, specially iii the presence
of covariates. A new approach to the representation of partitions and their interpretation
is also given. Applications jo Bayesian density estimation, linear
regression models and multivariate regression modeis for binary response are included. PFCHA-Becas Doctor en Estadística 116p. PFCHA-Becas TERMINADA