Abstract
The classical relationship between the spectral decomposition of a covariance matrix and the estimation of its principal components is utilized in obtaining robust covariance matrix estimates from robust estimates of the principal components, based on L1 formulations. The performance of these estimates is studied using some problematical data sets discussed in the literature.
Original language | English (US) |
---|---|
Pages (from-to) | 305-319 |
Number of pages | 15 |
Journal | Computational Statistics and Data Analysis |
Volume | 5 |
Issue number | 4 |
DOIs | |
State | Published - Sep 1987 |
Keywords
- Covariance matrix estimation
- L norm
- Linear and quadratic programming
- Principal component estimation
- Robust estimation