Abstract:
|
Susanne Rässler was among the pioneers promoting multiple imputation applications beyond the traditional nonresponse context. Two of these applications will be presented in this talk. Synthetic data are an elegant way to provide a database that preserves most of the information relevant to the analyst while ensuring confidentiality. Closely related to standard multiple imputation, the key difference lies in the fact that sensitive values instead of missing values are replaced by imputations. In this part of the talk, we will provide an overview of the synthetic data projects which have been conducted at the Institute for Employment Research (IAB) in recent years. We will also discuss Susanne’s pivotal role in initiating this research during her time at the IAB. Split questionnaire survey designs (SQSD)are deliberately created missing-by-design patterns. The basic idea is that shorter interviews yield smaller response burden. This, in turn, reduces measurement error as well as unit-nonresponse. We propose a genetic algorithm for finding an efficient SQSD that preserves information to the highest degree possible, and we compare the results to alternative split questionnaire design.
|