Using stochastic prior information in consistent estimation of regression coefficients in replicated measurement error model

A replicated ultrastructural measurement error regression model is considered where both predictor and response variables are observed with error. Availability of some prior information regarding regression coefficients in the form of stochastic linear restrictions is assumed. Using this prior information, three classes of consistent estimators of regression coefficients are proposed. A two-stage procedure is discussed to obtain feasible version of these Stochastically Restricted estimators. The asymptotic properties of the proposed estimators are studied. No distributional assumption is imposed on any random component of the model. Monte Carlo simulations study is performed to assess the effect of sample size, replicates and non-normality on the estimators. The methods are illustrated using real economic data.