New methods for generating significance levels from multiply-imputed data [Elektronische Ressource] / vorgelegt von Christine Licht

New methods for generating significancelevels from multiply-imputed dataDissertationzur Erlangung des akademischen Gradeseines Doktors der Sozial- und Wirtschaftswissenschaften(Dr. rer. pol.)an der Fakultat¨ Sozial- und Wder Otto-Friedrich-Universitat¨ Bambergvorgelegt vonChristine Lichtaus ApoldaBamberg, Oktober 2010Date of the defence: 2010-12-10Prof. Dr. Susanne Raessler (1st referee)Prof. Dr. Donald B. Rubin (2nd referee)AcknowledgmentsFirst, I would like to thank my advisors Susanne Rassler¨ and Donald B. Rubin fortheir support. Susanne Rassler¨ introduced me to missing-data problems and in-vited me to join the world of multiple imputation. She attended my first steps inthis field and prepared me meeting and finally doing research with Donald B. Ru-bin, the ”father” of multiple imputation. He also is the ”father” of this thesis, sincehis incredible previous and current ideas and the close co-operation with him arethe fundament of this thesis. I would like to thank him for the uncountable lessonsin multiple imputation theory, for his patience, when he answered all my moreor less smart questions, for inviting me to do research at the Harvard Statisticsdepartment, and in general for the whole support of this thesis.I am very greatful to Holger Aust, who supported me whenever it was neededand beyond. He motivated me in difficult times, when no solution of the trickyproblems was in sight.