On Tue, Aug 10, 2010 at 9:14 PM, Howie Fung hfung@wikimedia.org wrote:
I'd love to do some sort of editing frequency/time clustering. One thing that might be interesting is to take a cohort of users that started editing in, say Jan 2008, and map out that cohort's edit frequency by month from Jan 2008-present and see if meaningful clusters emerge. And then repeat for different start dates.
That would be fascinating.
users self-identified as "not having left", but I'm very curious as to whether they actually came back. I will try to get that stat. But on a
This points to two different useful characteristics. * people who consider themselves Wikipedians (and what that entails) * people who edit regularly (say, a few times a month)
The latter are contributing steadily to the projects (and might include unflagged bots, and other editors who would not consider themselves wikipedians). The former are people who are available to help in some way, or perhaps taking part off-wiki.
(I apologize to any bots in the audience who consider themselves Wikipedians. I'm looking at you, Emijrpbot.)
more general level, it would be great to know "for user has been editing for x months, with an average of x edits/month, if s/he stops editing for 3 months, there's an x% chance that s/he is not coming back."
Or, if we don't know what 'coming back' means, replace "is not coming back" with "will not edit more than z times in the following year"
SJ