Estimation for a unit level model with measurement errors


Accepted presentations cancelled by authors

Author: Laura Dumitrescu (University at Wellington, New Zealand)

Mixed effects regression models are widely used in small area estimation for linking areas and borrowing strength across domains but when the auxiliary information used in these is measured with error, the resulting estimators might be largely affected if the measurement error is ignored. Measurement error models are, in general, not identifiable and several methods have been proposed to obtain consistent estimators. Within a frequentist framework we construct a predictor that accounts for sampling variability in the auxiliary information, investigate its performance through simulated data and discuss an application to real data obtained from an election study.