Fractional Hot Deck Imputation for Robust Parameter Estimation Under Missing at Random

• Activity Number: 669
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Shu Yang*+ and Jae-Kwang Kim
• Companies: Iowa State University and Iowa State University
• Keywords: Robust estimation ; Multiple imputation ; Fractional imputation ; Hot deck imputation ; jackknife variance estimation

Abstract

Parametric fractional imputation (PFI), proposed by Kim (2011), is a tool for general purpose parameter estimation under missing data. We propose a fractional hot deck imputation (FHDI) which is more robust than PFI or multiple imputation. In the proposed method, the imputed values are chosen from the set of responses and are assigned with proper fractional weights. The weights are then adjusted to meet some calibration conditions, which make the resulting FHDI estimator fully efficient. Even though FHDI also need model assumptions, by the way of constructing the fractional weights, the consistency of the FHDI estimator is not affected by certain level of departures from the true model. Variance estimation from linearization is also discussed. A limited simulation study compares the proposed methods with other existing methods.