Efficient Multiply Robust Imputation Procedures in the Presence of Influential Units

• Activity Number: 245 - New Innovations and Challenges in Survey Sampling
• Type: Invited
• Date/Time: Tuesday, August 4, 2020 : 1:00 PM to 2:50 PM
• Sponsor: Canadian Statistical Sciences Institute
• Author(s): Sixia Chen* and David Haziza and Victoire Michal
• Companies: University of Oklahoma Health Sciences Center and Université de Montréal and McGill University
• Keywords: Influential units, Imputation, Multiply robust

Abstract

Item nonresponse is a common issue in surveys. To reduce the bias of unadjusted estimators, it is common practice to impute the missing values, leading to the creation of a completed data file. In practice, one must also face the problem of influential units in the sample, which make the commonly used estimators of population totals/means very unstable. To reduce the impact of influential units, we develop a robust version of multiply robust estimators using the conditional bias of a unit. The latter is a measure of influence of a unit that accounts for both sampling and nonresponse. We will present the results of a simulation study to show the benefits of the proposed method in terms of bias and efficiency.