Sample surveys typically gather information on a sample of units from a finite population and assign survey weights to the sampled units. Surveys frequently have missing values for some variables for some units. Imputation is widely used in sample surveys as a method of handling the missing data problem.

We provide a new imputation procedure using empirical likelihood to provide an easy-to-use data set for general purpose and keep the desirable properties of deterministic imputation method under fractional imputation. The imputed estimator constructed by the proposed procedure is called the Fractional Deterministic mputation (FDI) estimator. The construction of the FDI method is discussed in detail in order to describe the general proposed procedure. In addition, a computationally efficient variance estimator is given that permits the construction of general purpose replicates for variance estimation. In order to deal with multivariate missing data, the proposed imputation method can be extended via calibration to match results obtained using maximum likelihood estimation for some parameters under multivariate normal distribution model.

Finally, we address an issue common to several weight adjustment methods; namely, the issue of highly variable or even negative weights. Adapting an existing algorithm, our modification provides solutions with positive weights within a bounded interval. The combined contributions of this dissertation extend methods of imputation for missing values in important ways. Methods can be used with both complex survey and other data.