Research talk:Active editor spike 2015

Work log

[ refresh list ]

• Active editor spike 2015/Work log/2015-04-08
• Active editor spike 2015/Work log/2015-03-22
• Active editor spike 2015/Work log/2015-03-15

January

It's normal to have a spike in January, isn't it? Looking at stats:EN/TablesWikimediaAllProjects_AllMonths.htm, a 9k spike is not unprecedented. --Nemo 22:01, 11 March 2015 (UTC)

Indeed. I haven't had a chance to take a look at it yet, but User:Dario (WMF) says that there seems to be some unseasonably high numbers happening when he did a component extraction. --Halfak (WMF) (talk) 22:15, 11 March 2015 (UTC)

I generated time series decomposition plots for a number of large wikis, see below. This is based on active editor data generated from the databases (see the db definition used here).The main takeaways for English Wikipedia are the following:

• January 2015 was the strongest month since June 2013
• December 2014 was the first December ever (i.e. since 2001) not showing the typical seasonal drop (in fact in 2014 we had more active editors in December than in November, which is unusual).
• when you remove the seasonal component, the resulting trend indicates that in 2014 the decline slowed down and flattened;

I'd like to dig into the mobile vs desktop breakdown of active editors. The non-seasonal figures in the winter combined with the fact that the trend saw a possible inflection point in 2014 (a year that saw unexpectedly strong organic increase in mobile new active editors) make me guess that mobile might be the culprit here --DarTar (talk) 18:56, 15 March 2015 (UTC)

• 
  German Wikipedia
• 
  English Wikipedia
• 
  Spanish Wikipedia
• 
  French Wikipedia
• 
  Italian Wikipedia
• 
  Dutch Wikipedia
• 
  Russian Wikipedia

Breakdown by wiki size

In checking the graphs again I was quite struck by the breakdown by wiki, especially the graph for small wikis: it shows that in 2013 there was a drop from some 15k to some 10k "regularly active" editors, while all the other groups are rather constant. So I wonder how the breakdown is made: do you first deduplicate usernames and remove bots, then assign to a large/medium/small wiki set (in this case it's particularly important how you assign users to one group or another, maybe there is a bias towards big wikis?), or do you first assign and then deduplicate (in this case I'd see the graphs would not add up to the total one, but too hard to check manually)? --Nemo 07:21, 24 August 2015 (UTC)

July

Heh, it seems the trend was already interrupted, there is a decrease in YoY at stats:EN/TablesWikipediaZZ.htm#editor_activity_levels. Published today I see, thanks Erik. :) --Nemo 07:54, 25 August 2015 (UTC)