Timeline for Should one remove highly correlated variables before doing PCA?
Current License: CC BY-SA 3.0
4 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 23, 2013 at 18:06 | comment | added | whuber♦ | That's pretty close, Mohammad. The reality is that a set of correlated variables might "load" onto several principal components (eigenvectors), so including many variables from such a set will differentially weight several eigenvectors--and thereby change the directions of all eigenvectors, too. | |
| Feb 23, 2013 at 5:18 | comment | added | Spacey | @whuber Yes...yes I think I get it now. So essentially, more correlated variables will over-emphasize particular eigenvectors, (directions), and if there are many correlated variables, then there would be so many more overemphasized 'fake' directions, that drown out an 'original' eigenvector/direction that would have otherwise been easily seen. Am I understanding you right? | |
| Feb 21, 2013 at 22:02 | comment | added | whuber♦ | Did you read the hint by @ttnphns? PCA pays attention not just to the eigenvectors (which you discuss) but also to the eigenvalues (which you ignore). This is critical, because the eigenvalues are used to determine which components to retain and which ones to drop. | |
| Feb 21, 2013 at 19:31 | history | answered | Spacey | CC BY-SA 3.0 |