Synthetic graphical modelling of binary panel data subject to non-response

The present paper applies synthetic graphical modelling to the binary panel data of the Muscatine Coronary Risk Factor Study, where the observations are subject to non-response. The methodology combines the techniques of the graphical and generalized linear models. The purpose is to demonstrate the various aspects of the modelling approach than to obtain accurate empirical knowledge on the matter dealt with. We emphasize the flexibility, interpretation, and sensitivity (or robustness) of the models proposed.