Mira, Antonietta Tenconi, Paolo 2008-12-03 The main purpose of the thesis is the prediction of default events for small and medium size companies, by developing useful techniques able to cope with some issues related to anomalies often present in this kind of data, such as rare events and aberrant observations. To treat rare events the Bayesian paradigm is used through Markov Chain Monte Carlo techniques, also adopting the zero variance principle to reduce the variance of MCMC estimators. To deal with extreme observations the weighted likelihood approach is used in the context of logistic regression. application/pdf https://n2t.net/ark:/12658/srd1318133 https://susi.usi.ch/global/documents/318133 https://susi.usi.ch/documents/318133/files/2008ECO008.pdf eng info:eu-repo/semantics/altIdentifier/urn/urn:nbn:ch:rero-006-108840 info:eu-repo/semantics/altIdentifier/ark/12658/srd1318133 info:eu-repo/semantics/openAccess License undefined Credit risk Rare events Weighted likelihood Bayesian inference Markov chain Monte Carlo Variance reduction info:eu-repo/classification/udc/33 Statistical analysis for credit risk modelling http://purl.org/coar/resource_type/c_db06