IMPUTATION OF INDIVIDUAL VALUES OF A VARIABLE USING PRODUCT PREDICTORS

Abstract

Missing data is a common problem present in almost every data collection. It is particularly important in sample survey
research. The existence of missing observations (non-response ) is solved in many sample surveys using some technique of
imputation. They permit replacing the missing data. Several imputation techniques have been developed in the specialized
literature. The capacity of them for predicting the mean or a total is the token for their evaluation. This paper presents an
imputation rule with the main goal of predicting individual values. An auxiliary variable X is known for all the units and a
product type predictor is developed for predicting the value of variable of interest in each non-respondent. The
unbiasedness of the predictions is derived and the Mean Squared Errors (MSE`s). Some conclusions are pointed out.