Partial cross correlations and its applications to financial time series clustering
Cross correlation is widely used in many areas to measure mutual influence between variables. However, most cross correlation metrics measure the mutual information between a pair of variables, when in fact these can also be correlated with a third variable. In this FYP, we define a partial cross co...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
2015
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/62110 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Cross correlation is widely used in many areas to measure mutual influence between variables. However, most cross correlation metrics measure the mutual information between a pair of variables, when in fact these can also be correlated with a third variable. In this FYP, we define a partial cross correlation that measures the mutual information that is exclusively between a pair of variables. We apply this partial correlation metric to the cross section of S&P 500 stocks, and test for systematic differences between the linear and partial correlations. We also address the question of whether the partial correlation can be distorted because an incomplete cross section of variables is used. By computing the partial correlations within toy networks, we realized that the distortions carry signatures of the network topologies. We further explain how these signatures can be understood within the framework of the controllability of complex networks. |
|---|