Models of Association Versus Causal Models for Contingency Tables
Radu Tunaru · Journal of the Royal Statistical Society Series D (The Statistician) · 2001
The paper compares graphical association models and graphical chain models for contingency tables summarizing road accident data. It is shown that, for the same set of data, the two types of analysis may give different answers. It is described how to use collapsibility to reduce the dimensionality of the data without having problems with Simpson's paradox