Can I skip calculating metrics with incomplete information without throwing an error?

[gotoariel]

I have a relatively small dataset with large itemsets up to size 70. I'm trying to use fpmax because otherwise there's way too many subsets of itemsets overlapping. However as soon as I switch from apriori to fpmax, I get KeyErrors because (I think) of incomplete support information.

I'm wondering if there's a way I can just skip the offending itemsets instead of it stopping the whole process. I'm not sure how useful support_only will be if I can't sort by confidence or conviction at the end. Actually, what does association_rules with support_only do that's different from the support already generated in the frequent_itemset dataframe?

[rasbt]

Yeah, I agree that having an option to skip the offending itemsets would be ideal, but this is not implemented. I think the best workaround at the moment is using support_only=True and then computing the other metrics manually afterwards