Advanced Bayesian Inference, Spring 2022

[mlwilcox]

Attached are my updated code and results. I'm also attaching my data file again, and images of the 25 structures that I'm testing in the code.

The code was updated to compute and output the fold-changes for each possible pair of structures for BGe and BDe score generated using bnlearn and BiDAG. A fold-change >1 indicates that the second structure in the pair is more likely than the first structure; while a fold-change <1 indicates that the first structure is more likely. I then compared the fold-changes for BGe generated using bnlearn to the fold-changes for BGe using BiDAG, and identified discrepant results (i.e., when the fold-change from bnlearn was <1 and that from BiDAG was >1, or vice versa). Similarly, I identified discrepancies for BDe scores generated using bnlearn vs. BiDAG. The discrepancies are highlighted in the Excel file ending in "Comp25Strcts_mlw". For BGe, there were 19 (6.33%) discrepant pairwise comparisons. For BDe, there were 55 (18.33%) discrepant pairwise comparisons. Also, the log scores were the same for all i-equivalent structures for both BGe and BDe using both bnlearn and BiDAG (see output file ending in "Scores25Strcts").

In addition to updating the above, I have started drafting my final paper.