CMStatistics 2016: Start Registration
View Submission - CFE
Title: Cross-sectional multivariate ordinal regression models with an application in credit risk Authors:  Rainer Hirk - Vienna University of Economics and Business (Austria) [presenting]
Laura Vana - WU Wirtschaftsuniversitaet Wien (Austria)
Kurt Hornik - WU Wirtschaftsuniversitaet Wien (Austria)
Abstract: Two different approaches are investigated and extended for modelling multivariate categorical data. Both approaches augment cumulative link mixed models to a multivariate framework where several ordinal response variables are modelled jointly. The first approach uses the correlation structure of the error terms in order to capture the correlation among the different response variables. Composite maximum likelihood techniques in combination with a Metropolis-Hastings algorithm for optional random effects are used as estimation procedures. The second approach constitutes of a model with random effects and uncorrelated error terms where different estimation procedures like Laplace approximation and Gauss-Hermite quadrature are applied. All these procedures are implemented and analyzed in a simulation study. In addition, the models are fitted to real data with an application in credit risk.