Unit:
Κατεύθυνση Στατιστική και Επιχειρησιακή ΈρευναLibrary of the School of Science
Author:
Paraskevopoulou Chrysoula
Supervisors info:
Λουκία Μελιγκοτσίδου, Επίκουρη Καθηγήτρια, Τμήμα Μαθηματικών, Εθνικό Καποδιστριακό Πανεπιστήμιο Αθηνών
Original Title:
Bayesian Variable Selection for Normal and Generalized Linear Models
Translated title:
Bayesian Variable Selection for Normal and Generalized Linear Models
Summary:
The present thesis is about conducting inference for Normal and Generalized Linear Models in the framework of Bayesian Statistics. Our interest mainly lies on the search of more parsimonious models, using the method called "Stochastic Search Variable Selection" (SSVS). The method is thoroughly analyzed and applicated in the models mentioned above. Especially, regarding its application in the case of the Logistic Regression Model, Gamerman's Independence Sampler algorithm is proposed as a way of fully automizing the algorithmic process. An analysis of the theoretical background of the Method is provided , the corresponding algorithmic schemes are outlined, whereas applications in simulated data are also included.
Main subject category:
Science
Keywords:
Bayesian Statistics, Variable Selection, Stochastic Search Variable Selection (SSVS), Logit Model, Probit Model, Gamerman's Independence Sampler
