嘉宾:华磊 博士(Northern Illinois University)
地点:管理科研楼EMBA第五教室
时间:2016年6月17日(周五)上午11点
主办:管理学院统计与金融系、安徽省金融信息重点实验室
摘要: Comonotonic latent variables are introduced into general factor models, to allow non-linear transformations of latent factors, so that a variety of multivariate dependence structures can be captured. By employing a factor matrix of latent variables, we propose a unified representation that embraces several existing multivariate models. The latent variables within each column of the factor matrix are comonotonic, and the dependence among those columns can be modeled by dependence models that are of much lower dimension. The structure allows for multiple sets of comonotonic latent variables, as well as dependence clusters that may be overlapping. Numerical methods for estimation with the resulting copula models are studied. I will also demonstrate two applications with the proposed comonotonic factor model: one is for modeling dependence between human body components, and the other is for modeling cross-sectional dependence between many financial time series. The talk is based on a joint work with Professor Harry Joe.