A Test for Correlation between Signal and Noise within the Errors in Variables Model

When testing for measurement error, the vector of contrasts is the difference between the OLS and IV solutions. When testing for correlated measurement error, the OLS estimator must be replaced by a statistic which achieves consistency under the null hypothesis of uncorrelated measurement error. We propose one such estimator when one amongst several regressors is assumed to be measured with noise, and derive the related Hausman-type test.

MoreLess

Year of publication:	2002-04
Authors:	NAGA, Ramses H. ABUL
Institutions:	Départment d'économétrie et d'économie politique (DEEP), Faculté des Hautes Études Commerciales (HEC)
Subject:	errors in variables \| correlated measurement error \| consistent adjusted least squares \| Hausman tests

More details

Extent:	application/pdf
Series:	Cahiers de Recherches Economiques du Département d'Econométrie et d'Economie politique (DEEP).
Type of publication:	Book / Working Paper
Language:	English
Notes:	8 pages
Classification:	C1 - Econometric and Statistical Methods: General ; C4 - Econometric and Statistical Methods: Special Topics
Source:	RePEc - Research Papers in Economics

Persistent link: https://www.econbiz.de/10005650163