Open johnbradley opened 2 years ago
@hlapp suggested the following: After removing terms with 0 frequency check Jaccard similarity. If any terms have a Jaccard similarity of 0 raise an error instead. It would also be good to show a warning when removing any rows as users might not expect that this is happening.
After removing them, not before.
The resnik_similarity() function removes terms with 0 frequency. It is possible that this logic might remove the only common subsumer between two terms which would produce invalid results. Prevent this problem from occurring.
See https://github.com/phenoscape/rphenoscape/issues/235#issuecomment-917050436 for more details.