Errors in Variables for Numerical Analysts

Numerical analysts and statisticians are both concerned with errors in their least squares and regression matrices. However, their respective errors have different sources and pose different problems. Consequently, the two camps speak different languages, even when their algorithms are mathematically the same. In this paper we consider some measurement error models and derive the basic formulas in a way that should be accessible to numerical analysts with a minimal background in probability theory. This report is available by anonymous ftp from thales.cs.umd.edu in the directory pub/reports. y Department of Computer Science and Institute for Advanced Computer Studies, University of Maryland, College Park, MD 20742. This work is supported in part by the National Science Foundation under grant CCR 95503126. Errors in Variables for Numerical Analysts G. W. Stewart ABSTRACT Numerical analysts and statisticians are both concerned with errors in their least squares and regression mat..