search

kurtawirth / botscan

R tool for scanning Twitter for bot activity on a conversation level.
GNU General Public License v3.0
10 stars 6 forks source link

Check user-level calculation #19

Closed ryantmoore closed 5 years ago

ryantmoore commented 5 years ago

As of 9a94bb3, the user-level proportion defined in botscan.R at line 99 divides the number of bots identified in the full list of usernames by the number of unique usernames. I propose that this should be the number of unique bots identified divided by the number of unique usernames.

Proposed fix:

Move line 86, used only in the calculation above, 

nbots <- sum(df_userbots$cap.universal > threshold)

within the if(user_level){} and structure as 

nbots <- sum(
    (overallscreennames %in% botscreennames) &
    (overallscreennames_score > threshold)
    )
kurtawirth commented 5 years ago

This has been completed as of commit #a12c18044401775da6a7337c4438070b4fc37d25. Closing issue.

ryantmoore commented 5 years ago

From lines 115-6, can you verify that

  1. (tweets$screen_name %in% bots$user.screen_name) will identify unique instances of tweets$screen_name? That is, that there are no duplicates in tweets$screen_name.
  2. data frames tweets and df_userbots have the same observations in the same order?

If either is not true, I think we need to refix the calculation.

kurtawirth commented 5 years ago

With recent changes, this is handled differently and is now accurate.

ryantmoore commented 5 years ago

Fixed in commit 97a4a1926b2337cafc94e67fbc716f44f73ff5ad