|
Activity Number:
|
445
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #300562 |
|
Title:
|
Destructive Effect of the Noise in Principal Component Analysis with Application to Ozone Pollution
|
|
Author(s):
|
Katerina Tsakiri*+ and Igor G. Zurbenko
|
|
Companies:
|
University at Albany and University at Albany
|
|
Address:
|
1400 Washington Avenue, Albany, NY, 12222,
|
|
Keywords:
|
multivariate analysis ; principal component ; eigenvector
|
|
Abstract:
|
We consider statistical multivariate analysis of m normally distributed random variables when they are affected by the independent noise. Principal components, determined from joint distribution of original sample affected by noise, can be essentially different in comparison with principal components determined from original sample. Examples are provided. Asymptotic properties of covariance matrix affected by noise are investigated when the value of the noise is diminishing. When the main eigenvalues of original covariance matrix are all distinguishable, the effect of small noise proved to be neglectable. Simulation study is provided to determine critical levels of nearly correct determination of principal components in the noisy case. Result is applied to the analysis of the main factor in air pollution data.
|
