Analyzing Methods of Imputation

• Activity Number: 189
• Type: Invited
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: IMS
• Author(s): Scott Daniel Crawford*+
• Companies: Texas A&M University
• Keywords: semi-parametric ; missing at random ; efficient estimation

Abstract

There are many methods in the literature for imputing missing responses in semi-parametric regression, but not many of these articles address efficient estimation. Recently, Muller (2009) suggests an efficient approach in a nonlinear regression setting. The drawback of that approach is that it requires constructing an efficient estimator for the finite dimensional parameter, which can be quite involved. Under the assumption of responses missing at random I will investigate full imputation methods for several scenarios, using the efficient method suggested by Muller (2009) as well as other possibly inefficient approaches. I will analyze the variance of the estimators, both theoretically and with simulations.