Investigation on Life Satisfaction Through (Stratified) Chain Regression Graph Models

The study of marginal and/or conditional relationships among a set of categorical variables is widely investigated in the literature. In this work we focus on Chain Graph Models combined with the Hierarchical Multinomial Marginal Models and we improve the framework in order to take into account the context-specic independencies that are particular conditional independencies holding only for certain values of the conditioning variables. Letting the role of the variables to be purely explicative, purely response or mixed, in particular, we consider the (Stratied) Chain Regression Graph Model. A social application on life satisfaction is provided in order to investigate how the satisfaction of the interviewees' life can be affected by individual characteristics and personal achievement and, at the same time, how the personal aspects can affect the educational level and the working position.