Abstract: We consider a method of defining trimmed estimators of coefficients in an errors-in-variables model using the trimmed least squares estimators suggested by Koenker and Bassett (1978). The resultant estimators are consistent and asymptotically normal. In terms of the asymptotic relative efficiency, these trimmed estimators are more efficient than the traditional ones when the regression error in the errors-in-variables model has a heavy tailed distribution. A lower bound for the asymptotic relative efficiency is also established under some assumptions.
Key words and phrases: Errors-in-variables, trimmed least squares estimator, asymptotic relative efficiency.