This is a research contribution in the area of graphical model development and determination. The author presents an interesting approach to Bayesian graphical model determination, based on decision theory axioms. Although similar work is not often found in literature reviews, an extended overview is provided, also presenting the general concepts of graphical models.

An important aspect of the proposed approach is that it demonstrates the importance of considering both decomposable and non-decomposable models in graphical model determination. The author documents the proposed unified approach to graphical model determination through a considerable number of mathematical equations and definitions. The application of the introduced methodology to multinomial and multinormal data is carried out using both real and simulated data sets. The reader must be quite familiar with the Bayesian theory and methods for model determination in order to be able to follow the approach. Many references are provided, however, which allow the reader to obtain more information on the methods employed, and on related approaches.

The author finishes the paper with suggestions for future development of the current approach, to be extended to handle situations with missing data, an issue that has not yet been sufficiently addressed in any paper concerning Bayesian graphical model determination. Overall, the paper is well written, although it will not be easy to follow for anyone without a background in this specific field.