Three Definitions of Best Linear Approximation
Abstract
Three definitions of best (in the least squares sense) linear approximation to given data points are presented. The relationships between these three area discussed along with their relationship to basic statistics such as mean values, the covariance matrix, and the (linear) correlation coefficient. For each of the three definitions, and best line is solved in closed form in terms of the data centroid and the covariance matrix.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1976
- Accession Number
- ADA023501
Entities
People
- James J. Hurt