Skip to main content
3 events
when toggle format what by license comment
Aug 24 at 21:15 comment added Mauricio Mandujano M. Thank you! I think this solves the mystery. And for sure, data counterfactuals have to be carefully constructed. Though in my case it makes sense to use the full range above and below because that will be a "balanced grid" - counterfactually it'll weight what would have been the effect had this row entry had a 1 for all entries, then 2, and so on. This of course has its own assumptions (my DAG should be correct) but is better than making predictions tied to a specific sample that might be unbalanced.
Aug 24 at 21:02 vote accept Mauricio Mandujano M.
Aug 24 at 20:55 history answered EdM CC BY-SA 4.0