Search

Item nonresponse is often dealt with through imputation. Marginal imputation, which consists of treating separately each variable requiring imputation, generally leads to biased estimators of parameters (e.g., coefficients of correlation) measuring relationships between variables. Shao and Wang (2002) proposed a joint imputation procedure and showed that it leads to asymptotically unbiased estimators of coefficients of correlation. In this paper, we propose a modification of the Shao–Wang procedure, where initial imputed values obtained using this method, are modified so as to satisfy calibration constraints, which corresponds to MIVQUE estimators of model parameters. When the bivariate distribution of the variables being imputed is symmetric or exhibits a low degree of asymmetry, the proposed procedure is shown to be significantly more efficient than the Shao–Wang procedure in terms of mean square error. Results from a simulation study supports our findings. [ABSTRACT FROM AUTHOR]

Abstract: The problem of imputing missing observations under the linear regression model is considered. It is assumed that observations are missing at random and all the observations on the auxiliary or independent variables are available. Estimates of the regression parameters based on singly and multiply imputed values are given. Jackknife as well as bootstrap estimates of the variance of the singly imputed estimator of the regression parameters are given. These estimators are shown to be consistent estimators. The asymptotic distributions of the imputed estimators are also given to obtain interval estimates of the parameters of interest. These interval estimates are then compared with the interval estimates obtained from multiple imputation. It is shown that singly imputed estimators perform at least as good as multiply imputed estimators. A new nonparametric multiply imputed estimator is proposed and shown to perform as good as a multiply imputed estimator under normality. The singly imputed estimator, however, still remains at least as good as a multiply imputed estimator. [Copyright &y& Elsevier]