|
Abstract:
|
I have known Susanne since the first semester of my Bachelor studies. I was immediately excited about her kind nature and her passion for statistics. She had a massive impact on my professional life and without her I would not be where I am today. I was quickly enthusiastic about her main research focus - the handling of missing data by multiple imputation. Since then the appropriate treatment of missing values has been one of my greatest research interests. Susanne finally became my PhD supervisor. In my PhD thesis, I deal with imputation methods for non-ignorable missing values in the context of survey data which consist of data features such as repeated measurements and hierarchical structures. Standard applications of multiple imputation (MI) techniques assume that the data are Missing at Random (MAR). However, in many situations it seems very realistic that the missing values follow a Missing Not at Random (MNAR) mechanism. In this case, usual implementations of MI may lead to biased estimates. I will present different approaches to handle incomplete survey data that are supposed to be MNAR and will demonstrate their performance in different simulation scenarios.
|
