Consistent Estimation of Linear and Nonlinear Errors-in-Variables Models with Validation Information

Abstract

Consistent methods for the estimation of linear and nonlinear regression models with general measurement errors in variables in the presence of validation data information are introduced. The estimation procedures do not rely on any specific specification of auxiliary structural measurement error equations and are robust against any specification of measurement error equations. the estimation procedures can be applied even to situations where validation data alone does not permit possible estimation of the model. When validation data is rich enough for consistent estimation of the model, survey data are still valuable in improving efficiency of the validation data estimates. The estimation procedures can be applied to models and data where direct bias corrections might not be feasible.