A new ratio-type imputation with random disturbance Original Research Article

In this study we consider the problem where two variables can have missing values not necessarily for the same units. We propose a class of imputed estimators by means of a ratio technique with random disturbance. Next we derive the asymptotically optimal estimator into the class. The simulation study shows that the proposed method produces an important increase in efficiency when parameters such as the median, quartiles and variance are estimated.