| |
 |
|
|
Science Forum Index » Environment Forum » Auditing the Auditors # 11
Page 1 of 1
|
| Author |
Message |
| David Ball |
 Posted: Wed Dec 17, 2003 10:40 pm |
|
|
|
Guest
|
M&M's question 11 is a bit odd....
11. What is the basis for inclusion of some tree ring sites within a
region in regional principal component calculations and others as
individual dataset components?
PCA is a technique for reducing the dimensionality of
multivariate datasets. If you can represent a given dataset,
accounting for a significant portion of the variance in the data in
the process, with a reduced set of principle components, you use PCA.
If, in a given region, you have certain series that cannot be
adequately described by the recovered PC's then it should not be
discarded. Essentially, a series that cannot be represented by the
PC's is an outlier. Such series have valuable information in them and
must be included, unless of course the goal is to remove outliers from
the study.
When you construct you correlation or covariance matrix,
depending on the data you are looking at, you may find that certain
variables have next to no contribution to the PC's. Removing them
doesn't affect the underlying PC's. Still, they may contain
information, so you retain them as individual series for completeness. |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Fri Aug 29, 2008 3:00 pm
|
|