rexford

On Wednesday, November 27, 2013, Erik Zachte wrote:

Isla,

Two different sets of wikistats reports use a slightly different definition of ’edit’. With one major difference.

1
The dump based reports (e.g. on articles, editors, etc) count every revision of a page stored in the database,
for articles which are perceived to be encyclopedic content or ‘content’ for short.

For most wikis only namespace 0 is such countable content.
A few wikis have extra namespaces deemed as content, which these are can be queried via an API call.
From these namespaces formally only pages containing an internal link are to be included (but in practice all are, as that info is not always available).

Again with the exception of redirects which are always excluded.

Then Wikistats reports list counts separately for registered users, anons and bots
See e.g. http://stats.wikimedia.org/EN/TablesWikipediaEN.htm#editor_activity_levels and other tables on that page

2
Our traffic reports, based on server logs, count any page submit to any wiki html page, no matter which namespace.
But the proportion of edits to non encyclopedic content is pretty small, so no big deal.

Here bots are filtered at all (assuming x+ edits per day signals a bot, taking few false positives for granted)
Major difference: These scripts do not know about logged-in status of person (so total edit count is larger).

As these reports look into geo-distribution this is well worth mentioning.

http://stats.wikimedia.org/wikimedia/squids/SquidReportPageEditsPerCountryOverview2013Q3.htm

In both occasions only a final submit of an edit is counted, not the opening of the page in edit mode, or previews.

(also resending response to Rupert to all)

Rupert,

The report you mention is actually to be replaced by a more precise one.
Numbers for that report were based on 1:1000 sampled log.

New reports for edits use 1:1 unsampled log and are hence fully precise.
And numbers are monthly edits.
See for example this quarterly report:

http://stats.wikimedia.org/wikimedia/squids/SquidReportPageEditsPerCountryOverview2013Q3.htm

Cheers,
Erik

PS
Unrelated: today I saw this nice quiz:
http://youdontknowafrica.com/?location=africa

From: Isla Haddow-Flood [mailto:islahf@africacentre.net]
Sent: Monday, November 25, 2013 11:58
To: rupert THURNER; wikimedia-gh@lists.wikimedia.org; wikimediake-l@lists.wikimedia.org; wikimediaza; Erik Zachte
Subject: Re: [Wikimedia ZA] page edits per country ...

Hi Rupert

Thank you for the statistics!

I have often wondered that question and perhaps it is something we should ask Erik Zachte who collates the stats (and is now cc’d in this email). However, I am also not sure what exactly how to define a ‘real’ edit (excuse my naïveté). Does it mean one not created by a bot? Or the length of the text contributed, or indeed the quality of text. I am not sure how one measures the difference and if there is a tool for doing so.

Erik, perhaps you could clarify for us?

Warmest
Isla

                  WikiAfrica
Isla Haddow-Flood
Project Manager: Africa

tel:      +27 21 418 3336
skype: islahaddow
twitter: @wikiafrica
et started: http://en.wikipedia.org/wiki/Wikipedia:WikiAfrica
en.wikipedia.org/wiki/User:Islahaddow

From: rupert THURNER <rupert.thurner@gmail.com>
Date: Mon, 25 Nov 2013 09:16:23 +0100
To: "wikimedia-gh@lists.wikimedia.org" <wikimedia-gh@lists.wikimedia.org>, "wikimediake-l@lists.wikimedia.org" <wikimediake-l@lists.wikimedia.org>, wikimediaza <wikimediaza@lists.wikimedia.org>
Subject: [Wikimedia ZA] page edits per country ...

hi

here page edits per month and country:
http://stats.wikimedia.org/wikimedia/squids/SquidReportPageEditsPerCountryOverview.htm

for Africa:
15'000 Egypt
10'000 south Africa
6'000 algeria
5'000 marocco
3'000 tunesia
2'000 kenya
1'000 angola
1'000 ghana

I am wondering how these numbers should be read. How many real edits does it mean?

rupert

_______________________________________________ WikimediaZA mailing list WikimediaZA@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikimediaza