We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Generalized Errors in Variables Regression.
- Authors
Casson, M. C.
- Abstract
When there is error in just one of the independent variables in a linear regression all the OLS slope estimators are inconsistent. Consistent estimators are well known for the case in which all non-constant independent variables have errors. The paper derives estimators for the mixed case in which some of the variables have errors and some do not. Estimators are particularly useful in studies of the permanent income hypothesis, in estimating production and investment functions where capital is measured subject to error and in generalizing Klein's method for the estimation of Koyck type distributed lags to rational distributed lags. An equation follows that in the special case of non-homogeneous regression and zero correlation between the stochastic term and the measurement error the slope estimator of the with-error variable is biased toward zero unless the true slope is zero, and the slope estimator of the ith error-free variable is biased as well unless the (m -2)th order partial correlation between the error-free and with-error observations is zero.
- Subjects
ERROR analysis in mathematics; MATHEMATICAL variables; ESTIMATION theory; REGRESSION analysis; STATISTICAL correlation; MATHEMATICAL statistics
- Publication
Review of Economic Studies, 1974, Vol 41, Issue 3, p347
- ISSN
0034-6527
- Publication type
Article
- DOI
10.2307/2296754