Jackknife Empirical Likelihood Method for Inference with Imputed Data

• Activity Number: 626
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Sixia Chen*+
• Companies: Iowa State University
• Keywords: Item nonresponse ; fractional imputation ; interval estimation ; regression imputation ; Poisson sampling

Abstract

We extend Jackknife empirical likelihood (JEL) method proposed by Jing etc. (2009) to make a likelihood-based inference with imputed data. Specifically, likelihood-ratio confidence intervals can be constructed using a chi-square limiting distribution if the empirical likelihood is applied to the jackknife pseudo values. The jackknife pseudo values are very close to the pseudo values for variance estimation discussed in Kim and Rao (2009). Using the empirical likelihood, we can incorporate additional auxiliary variables into estimation and improve its efficiency. The proposed method is applicable to some complex sampling designs such as stratified Poisson sampling under some regularity conditions. Extension to fractional imputation is also discussed. Simulation results are also presented.