On a symmetrized simulation extrapolation estimator in linear errors-in-variables models

Application of naive regression estimates in errors-in-variables models (EVM) suffers from a severe bias. The simulation extrapolation estimator (SIMEX) was introduced by Cook and Stefanski as a correction method modeling the dependence of error variance in the regressors and bias of the regression method. The symmetrized simulation extrapolation estimator (SYMEX), a generalization of SIMEX, allows to employ the symmetric structure of EVM. Relations of both SIMEX and SYMEX to total least squares are investigated. (C) 2003 Elsevier B.V. All rights reserved.