In this presentation, a pseudo empirical likelihood approach for survey data with nonresponse is introduced. We focus on efficient and doubly robust imputation for covariate-dependent missing responses, with application in combining imputation cells or datasets. Some theory, simulation, and example are presented.