I virtually always seem to get a TREE structure (i.e., branching, but no 'loops') when using a score-based method such as the Bayesian search (which uses the BDeu score). Is this typical for score-based methods ?
With the same data, the constraint-based methods (such as the PC algorithm variants) are much more likely to result in structures with 'loops' in the graph. This seems to occur with multiple data sets, and with many different alpha values in the PC algorithm.
Hence, I am wondering if this is a general characteristic of these two classes of algorithm, or is there something unique to my data (or the method I am using) which causes this to occur ?
Thank you in advance for any help you can offer !