Statistical Modeling Approaches for Behavioral and Medical Sciences

Behavior and abilities are arguably the most fundamental concepts in life. Hence, these concepts and their causes and consequences, are studied in many disciplines, such as economics, psychology, sociology, and health sciences. The research of behaviors and abilities is quickly changing: data sets are rapidly growing in size, both in terms of observations and variables. The increase in the sample size of studies, repeated measurements and complex multi-level designs necessitates the use of modern - often more complex and flexible - statistical methods. This thesis employs modern statistical models in the field of fundamental economic behavior and medical sciences with the aim to generate more reliable inferences and more accurate predictions than would have been the case with established methods. In the present thesis I will give two examples. In the first part of the thesis, I develop the Censored Mixture Model (CMM) to analyze risk behavior and show two applications. In the second part, I compare two machine learning approaches with various simpler commonly used baseline approaches in their performance to predict cognitive ability using brain morphology.