In this case, looking at simply the standard deviations is actually a little misleading.  I've posted the actual distributions to the page, which provides a different picture than the standard deviations would suggest.   There is a definite skew in the distribution of the data, especially for anonymous users.  Anonymous users are much more likely to give 5s while ratings from registered users appear to be more distributed.

http://www.mediawiki.org/wiki/Article_feedback/Public_Policy_Pilot/Early_Data#Comparing_Anon_Reviewers_to_Registered_Reviewers

Howie

On 9/30/10 1:54 AM, Finn Aarup Nielsen wrote:


On Thu, 30 Sep 2010, Magnus Manske wrote:

Just out of interest, if the maximum value is 5, and (for example) the
"neutral" mean is 3.6, how can the standard deviation be 1.82?
Wouldn't that allow values up to 5.42?

If that's an effect of extreme skewing, maybe the median would be
better suited to give a "common" value?

A data set [ 5, 5, 4, 3, 1 ] will have mean 3.6 and standard deviation 1.67. This is also above 5. There is skewness here. If you use the median (here 4) you wont get any decimals. I think the mean is fine.


cheers
Finn

___________________________________________________________________

         Finn Aarup Nielsen, DTU Informatics, Denmark
 Lundbeck Foundation Center for Integrated Molecular Brain Imaging
   http://www.imm.dtu.dk/~fn/      http://nru.dk/staff/fnielsen/
___________________________________________________________________



Cheers,
Magnus

On Thu, Sep 30, 2010 at 12:50 AM, Howie Fung <hfung@wikimedia.org> wrote:
Everyone,

As some of you may know, we launched an experimental Article Feedback
feature as part of the Public Policy Initiative last week.  The "Article
Feedback Tool" enables readers to quickly assessthe sourcing, completeness,
neutrality, and readability of a Wikipedia article on a five-point scale.
It is currently deployed on about 300 articles [1] in the area of Public
Policy on the English Wikipedia.  More details may be found on the blog post
[2] as well as the post on Foundation-l [3].

We've been capturing the ratings data and have some early analysis to share
around the types of ratings users are providing.  There are some interesting
differences between anonymous and registered users:

http://www.mediawiki.org/wiki/Article_feedback/Public_Policy_Pilot/Early_Data

The dump of the article-level data is also available [4] for those who are
interested.

If anyone would like to be involved in the ongoing research and evaluation
of this tool, please sign up on the Article Feedback Workgroup page. [5]

Howie

[1] http://en.wikipedia.org/wiki/Category:Article_Feedback_Pilot
[2]
http://blog.wikimedia.org/blog/2010/09/22/article-feedback-pilot-goes-live/
[3]
http://lists.wikimedia.org/pipermail/foundation-l/2010-September/061056.html
[4]
https://spreadsheets.google.com/ccc?key=0Aikdcg5HdSKbdFRhdUN1Rm1iZzB5dUdMUlY4YzAwNmc&hl=en#gid=0
[5]
http://www.mediawiki.org/wiki/Article_feedback/Public_Policy_Pilot/Workgroup

_______________________________________________
Wiki-research-l mailing list
Wiki-research-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wiki-research-l



_______________________________________________
Wiki-research-l mailing list
Wiki-research-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wiki-research-l 


_______________________________________________
Wiki-research-l mailing list
Wiki-research-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wiki-research-l