Navigation

Recent Publications

More...

Home

Modeling on Realized Covariance Matrix

patng324 - Posted on 21 December 2012

Project Description: 

This project aims at modeling daily realized covariance matrices which arise from the high-frequency financial data. The time-series model involves the noncentral wishart transition distribution. Maximum likelihood estimation will be used to fit the model which involves the likelihood function that does not have a closed-formed formula. The optimization problem encountered depends on the numerical solution of the likelihood function.

Researcher name: 
Ng Fo Chun
Researcher position: 
PhD student
Researcher department: 
Department of Statistics and Actuarial Science
Researcher email: 
patng324@hku.hk
Research Project Details
Project Duration: 
12/2012 to 08/2013
Project Significance: 
This project aims at illustrating that the dynamic of covariances among asset returns may be explained not only by time-varying scaling matrices which comes from heteroskedasticity of asset returns but also by time-varying noncentrality matrices which comes from autocorrelation between asset returns. Therefore, the model can be viewed as the dynamic bias correlation of realized covariance matrices.
Results Achieved: 
The result will be published as a research paper.
Remarks: 
Maximum likelihood estimation will be used to fit the model which involves the likelihood function that does not have a closed-formed formula, and so for its derivatives. And hence, the optimization problem encountered depends on the numerical solution of the likelihood function. Therefore, the computation involved in model estimation and forecasting will be quite heavy.