Research talk:Wikipedia article creation/Work log/Friday, November 22nd

Friday, November 22nd

Note that I'm posting this on Monday, Nov. 25th. I spent a bunch of time discussing my results on IRC, so I didn't post them here. Here's what was missed.

Article counts by account type. The proportion of surviving of articles are plotted by creation month with linear model fits overlayed.

The plot shows that the proportion of main namespace pages being kept has been rising slightly over time. This is unexpected -- especially in the case of page creators who have less than 24 hours of experience. I see a few explanations for this:

• AfC is taking over some of the load. AfC has authors create pages in Wikipedia_talk namespace and only move them once they are ready, so my methods would miss AfC created pages that did not pass review.
• Newcomers are actually less likely to create non-notable/spammy articles for some strange reason.
• The criteria for deletion are relaxing.

My next step is to replicate this analysis in non-English Wikis. --Halfak (WMF) (talk) 17:59, 25 November 2013 (UTC)