A Semiparametric Approach to Inference with Nonignorable Missing Data Using Surrogate Information

• Activity Number: 255
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Sixia Chen*+ and Jae-Kwang Kim
• Companies: Iowa State University and Iowa State University
• Keywords: Exponential tilting ; Not missing at random ; Nonparametric regression ; Response mechanism

Abstract

Parameter estimation with non-ignorable missing data is considered. To avoid the identifiability problems in the model, we assume that surrogate information is available throughout the sample so that the response mechanism can be modeled as a function of surrogate variable and other covariates available in the sample. We use an exponential tilting model to derive an imputation model from the model for the respondents. Exponential tilting model uses minimum assumptions and thus is robust.

In this paper, based on the exponential tilting model, we propose a semi-parametric estimation method of parameters with non-ignorable missing data. Variance estimation will be discussed and results from some simulation study will also be presented.