Proxy Variables to Common Factors and Parameter Estimation in Factor Copula Models

You are welcome to attend the following Statistics and Stochastic colloquium (part of the Colloquium Series of the Department of Mathematics and Statistics) at La Trobe University.

Speaker: Dr Pavel Krupskiy, University of Melbourne

Time & Date: 12:00 noon Thursday 4 April 2019

Venue: Room 310, Physical Sciences 2, La Trobe University, Melbourne Campus

Abstract:

Factor copula models assume observed variables are independent conditional on one or several unobserved factors. We show that, under some mild assumptions, proxy variables to the unobserved factors can be obtained from the observed variables. These proxy variables can help to select appropriate linking copulas in factor copula models and to get fast estimates of parameters of these copulas in high dimensions. We use simulation study to show that parameter estimates obtained using the proposed approach are very close to estimates obtained using the maximum likelihood approach which can be computationally demanding for big data sets. We apply the proposed approach to analyse a financial data set.