How the stated 0.0208% interaction density was calculated

#2
by abramsci - opened

There is a quite crucial proporion number mentioned first in Basic Statistic section and later in the table (Configurable Subsets section).

I understand that with the ongoing competition the organizer probably would be wise to not reveal how it is computed exactly, but I hope when the test set will be released it will be clarified. For now, I can only make two assumptions and provide my reasoning. Maybe it would be useful for some of my competitors or maybe the organizer will give some hints so I structured it in two parts. If answering question B will reveal too much, then I appreciate even a simple Yes or No for the question A 😄

  • Question A: 0.0208% = the stated interaction density - is it some percentage of total table binary part {like, dislike, share, bookmark, open_comments, click_on_author} interaction results?

  • Question B: I have two assumtions so the question is if either of these is correct?

      1. If it is across all 6 columns of the binary table (X / (6 * 40_774_024_903)) then X ~ 4_240_498_589
      1. If it is across logical OR {like ∨ share ∨ bookmark ∨ open_comments ∨ click_on_author} thenX ~ 848_099_718 (as we assume that X / 40_774_024_903 == 0.0208)

where X denotes total non-zero (true) interaction results.

Sign up or log in to comment