Estimation of regression parameters with arbitrary noise

Research output: Contribution to journalArticlepeer-review

Abstract

The problem of estimation of linear regression parameters with arbitrary (nonzero mean, correlated and even nonrandom) noise is considered. The essential assumption is that the inputs are random, centered and independent of the noise. Convergence and rates of convergence for stochastic approximation (standard and with averaging) and least squares estimators are investigated under these conditions. Applications and generalizations of the obtained results are discussed.
Original languageEnglish
Pages (from-to)18-29
Number of pages12
JournalMathematical Methods of Statistics
Volume2
StatePublished - Jan 1993

Fingerprint

Dive into the research topics of 'Estimation of regression parameters with arbitrary noise'. Together they form a unique fingerprint.

Cite this