Model-based recursive partitioning (MOB) provides a powerful synthesis between machine-learning inspired recursive partitioning methods and regression models. Hanna Birke extends this approach by allowing in addition for measurement error in covariates, as frequently occurring in biometric (or econometric) studies, for instance, when measuring blood pressure or caloric intake per day. After an introduction into the background, the extended methodology is developed in detail for the Cox model and the Weibull model, carefully implemented in R, and investigated in a comprehensive simulation study.

Contents

• MOB and Measurement Error Modelling
• Derivation of an Adjusted MOB Algorithm for Covariates Measured with Error for the Cox and Weibull Model
• Implementation of the Suggested Method for the Weibull Model in the Open-Source Programming Language R
• Simulation Study Showing the Performance of the Implemented Method Target Groups

• Researchers and students in the fields of statistics and cognate disciplines with interest in advanced modelling in combination with measurement error in covariates
• Data analysts of complex biometric or econometric studies with variables that are difficult to measure in practice The Author

Hanna Birke wrote her master thesis under the supervision of Prof. Dr. Thomas Augustin at the department of statistics of the LMU Munich and is currently working on her doctoral thesis.

Autorentext
Hanna Birke wrote her master thesis under the supervision of Prof. Dr. Thomas Augustin at the department of statistics of the LMU Munich and is currently working on her doctoral thesis.

Inhalt
MOB and Measurement Error Modelling.- Derivation of an Adjusted MOB Algorithm for Covariates Measured with Error for the Cox and Weibull Model.- Implementation of the Suggested Method for the Weibull Model in the Open-Source Programming Language R.- Simulation Study Showing the Performance of the Implemented Method.