Comparisons between Some Estimators in Functional Errors-in-Variables Regression Models.
Abstract
This report studies the functional errors-in-variables regression model. In the case of no equation error (all randomness due to measurement errors), the maximum likelihood estimator computed assuming normality is asymptotically better than the usual moments estimator, even if the errors are not normally distributed. For certain statistical problems such as randomized two group analysis of covariance, the least squares estimate is shown to be better than the aformentioned errors-in-variables methods for estimating certain important contrasts.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1982
- Accession Number
- ADA120391
Entities
People
- Paul P. Gallo
- Raymond J. Carroll
Organizations
- University of North Carolina at Chapel Hill