Dissertation
SPARSE MODEL SELECTION FOR HIGH-DIMENSIONAL VINE COPULAS VIA PENALIZED OPTIMIZATION
Doctor of Philosophy (PhD), Washington State University
01/2018
Handle:
https://hdl.handle.net/2376/100059
Abstract
The Vine Copula is a flexible graphical model to capture the dependence structure
for high-dimensional continuously-valued random variables. It has been shown to
be useful in many fields such as finance and risk management, especially when the
usual Gaussian assumptions do not hold. A couple of estimation methods for vine
copulas have been discussed. In this dissertation, we propose a method for model
selection of high-dimensional vine copulas with certain structures, such as the sparsity
(i.e., promote the number of independence copulas used). We start by the stepwise
estimation, which is a popular estimation method for the vine copula. Furthermore,
we introduced the zero-norm penalty functions that are used to identify the sparsity.
We report comparisons between our proposed method and the VineCopula package
on simulated data under different scenarios. At the end of the study, we also apply
the proposed method to real financial data (weekly return rate of stocks
Metrics
6 File views/ downloads
86 Record Views
Details
- Title
- SPARSE MODEL SELECTION FOR HIGH-DIMENSIONAL VINE COPULAS VIA PENALIZED OPTIMIZATION
- Creators
- Wei Li
- Contributors
- Hongbo Dong (Advisor)Haijun Li (Committee Member)Alan Genz (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Department of Mathematics and Statistics
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Number of pages
- 74
- Identifiers
- 99900581508001842
- Language
- English
- Resource Type
- Dissertation