Wikipedia:Statistics Department

This WikiProject is defunct.
Consider looking for related groups or ask for help at the Teahouse.

If you feel this group may be worth reviving, please discuss with related groups first. Feel free to change this tag if the parameters were changed in error.

Shortcuts

This project page concerns statistics about Wikipedia. For the WikiProject on the mathematical science of statistics, see WikiProject Statistics

This project, the Statistics Department, provides a space for contributors interested in statistics to discuss what to measure, when, and how.

If you would like to help, please add your name below and introduce yourself on the talk page. The to-do list below is just a start...

Scope

This WikiProject aims primarily to design, implement, and discuss the collection of statistics about Wikipedia content, metacontent, contributors, and visitors. We seek to better understand how people use Wikipedia and its community, and what is most useful to them. We also seek to explore new ways of streamlining the generation of timely statistics.

Participants

Please add your name here by adding ~~~

Opus meum 15:16, 23 March 2017 (UTC)

Pages

Research Questions

Contribution

Who contributes to Wikipedia, when during the day/week, and how often?
What causes sudden spikes in readers, contributors, vandals?
Are there patterns in the contributions? E.g. age, gender, race and nationality versus categories?
What motivated the top contributors? E.g. repute, reciprocity, altruism, relationships, roles? Free content, neutrality, software design, democracy, community, others?
How are the quality, validity and reliability of content maintained? By whom, and to what extent?
How does server load contribute to activity of users? in the hours/days after a slowdown?
Where (on Earth!) are the contributors? Are contributors to en.Wikipedia in English speaking countries, Spanish/Portuguese lang. contributors in Iberia or Latin America or elsewhere, German lang. contributors in German, Austria, Switz. or elsewhere, etc.

Promoting Readership/Consumption

Who reads Wikipedia articles, when?
What linkpaths do they follow through the site?
- What are common first pages visited?
- What are common pages visited from the Main Page?
How have changes to Recent Changes page and Main historically affected user clickthroughs from those pages?
How often do anonymous visitors/readers (or visitors from Google/Yahoo) visit pages like RC, Random, the Community Portal?
What are the readers' ratings of the quality or usefulness of each page?

Curtailing Mischief

How can we quantify vandalism? Trolling?
How many admins are online at a given time?
How does the # online relate to the amount of vandalism that takes place?
Are vandals deterred by quick response times?
How effective are bans and blocks? How often do vandals come back right away as anons or with another ip?
What is the average block length? How does the block length change from editors to IPs?
What is the median time-to-correction for acts of vandalism? (Recent study: Vandalism Survival.)

Processes

How do different people add content? <-- what does this mean (other than Edit This Page)? Elaboration needed.
Slow vs. fast contributors; people who write offline vs. online
- How many use offline editors, and upload in blocks?
How many people migrate content from other free repositories to WM sites?
- photos, text (to commons, source)

Methodology

This section should cover how the research data will be collected and analysed, and not Wikipedia context or processes (moved to above section).

Data Collection

Webalizer statistics
Add optional fields in every member's profile form for age, gender, race, nationality (perhaps with a privacy option - so system can collect data, but not visible to general public)
Polls for all in Community Portal
Surveys/Interviews of top contributors
- Constructs needed for different motivational factor
Toolserver

Data Analysis

Define & select uniform data structures and software (SPSS, SAS)
Define variables
- Outcome measures
Correlational designs
- t-tests
ANOVA/MANOVA (for correlational data)
- Post-hoc statistics (LSDs, Fischers)
Factor analysis
Non-parametric measures (Chi-Square)

Caveats?

Privacy
- Possible solution: Constrain to publicly available data; and, if private data must ever be used, absolutely no personally-identifiable info.
Consent to participate in certain surveys
- Possible solution: Avoid experimental setups, and avoid self-response surveys, as self-response is frequently difficult to gauge at times, as well. However, properly structured, anonymous polls that have pretty much no chance of "psychological trauma" or whatnot are probably safe :P
Feedback effects of certain metrics (edit #) via social loops (people editing for the sake of edit count)
- Possible solution/offset: Effect interactions betw. edit count/other factors; analysis of random sample of RfA fails vs. successes and method of analyzing primary rationale of voters?

References

Basic Statistics

Results

Statistics Scoreboard
Metric	Current Value
Users	55,144,324
Admins	818
User/Admin Ratio	67413.6 users per admin
Edits	1,377,158,183
Pages	66,486,451
Edits per Article	20.71 edits per page