I've been taking an extended wiki-break during the last 6 weeks, but during that time I've also reflected on the state of Wikipedia. Rather than simply relying on my own observations, I started asking questions and went looking for statistical answers.
As many of you are probably aware, the statistics package hosted at http://stats.wikimedia.org/EN/TablesWikipediaEN.htm has not been updated in about a year due to the failure of all recent attempts to create a complete database dump for enwiki.
Given the lack of any recent official stats, I set out to generate my own using a dump of the Wikipedia log files and by systematically downloading (over many days) the history page contents for 100,000 articles.
My analysis from doing this gave me a big surprise. Since early this year, and for the first extended period in Wikipedia's history, the activity rate of the Wikipedia community has been declining. This can be seen in the rate of editing articles (-17%), the rate of new account registration (-25%), blocks (-30%), protections (-30%), uploads (-10%), article deletions (-25%), etc. Some exceptions are the article creation rate (+25%) and image deletions (+80%), but overall the community appears to be doing less now than it was 6 months ago.
Given Wikipedia's long history of nearly exponential growth, any decline in our editing rate is surprising, but I also find the sharpness of the turn-around to be remarkable. Though it may merely be coincidence, the timing of the change is near the breaking of the Essjay story in the news.
At the same time the frequency of reverts (as a fraction of all article edits) continues to increase, while administrators now devote more than 1/3 of their article space edits to making reverts.
For charts and additional information see: http://en.wikipedia.org/wiki/User:Dragons_flight/Log_analysis
I'm not going to launch into an extended discussion of what's wrong with Wikipedia, but I do think we need to be paying attention to these trends and taking steps to intervene where possible.
-Robert Rohde aka Dragons_flight
You're trying to make an accurate judgment based on 100k of articles from a 2 million article field? Don't insult our intelligence.
On 10/9/07, Robert Rohde rarohde@gmail.com wrote:
I've been taking an extended wiki-break during the last 6 weeks, but during that time I've also reflected on the state of Wikipedia. Rather than simply relying on my own observations, I started asking questions and went looking for statistical answers.
As many of you are probably aware, the statistics package hosted at http://stats.wikimedia.org/EN/TablesWikipediaEN.htm has not been updated in about a year due to the failure of all recent attempts to create a complete database dump for enwiki.
Given the lack of any recent official stats, I set out to generate my own using a dump of the Wikipedia log files and by systematically downloading (over many days) the history page contents for 100,000 articles.
My analysis from doing this gave me a big surprise. Since early this year, and for the first extended period in Wikipedia's history, the activity rate of the Wikipedia community has been declining. This can be seen in the rate of editing articles (-17%), the rate of new account registration (-25%), blocks (-30%), protections (-30%), uploads (-10%), article deletions (-25%), etc. Some exceptions are the article creation rate (+25%) and image deletions (+80%), but overall the community appears to be doing less now than it was 6 months ago.
Given Wikipedia's long history of nearly exponential growth, any decline in our editing rate is surprising, but I also find the sharpness of the turn-around to be remarkable. Though it may merely be coincidence, the timing of the change is near the breaking of the Essjay story in the news.
At the same time the frequency of reverts (as a fraction of all article edits) continues to increase, while administrators now devote more than 1/3 of their article space edits to making reverts.
For charts and additional information see: http://en.wikipedia.org/wiki/User:Dragons_flight/Log_analysis
I'm not going to launch into an extended discussion of what's wrong with Wikipedia, but I do think we need to be paying attention to these trends and taking steps to intervene where possible.
-Robert Rohde aka Dragons_flight _______________________________________________ WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
On Wed, 2007-10-10 at 08:43 -0700, Steven Walling wrote:
You're trying to make an accurate judgment based on 100k of articles from a 2 million article field? Don't insult our intelligence.
Exactly what is new with taking a sample of the whole to come up with trends in statistics ?
Any concern should be whether 5% of 2 millions is a high enough sample (I would say so), and whether the samples is representative of the whole sample space (all the articles). The second I can't comment on as I do not know what those 100K articles are among other reasons.
KTC
The editing sample is based on a random number generator selection of articles (namespace=0, page_is_redirect is false) in the September 8th dump of the page table. So it excludes articles created in the last weeks of September, but is otherwise a random sample of everything in article space.
-Robert
On 10/10/07, Kwan Ting Chan ktc@ktchan.info wrote:
On Wed, 2007-10-10 at 08:43 -0700, Steven Walling wrote:
You're trying to make an accurate judgment based on 100k of articles
from a
2 million article field? Don't insult our intelligence.
Exactly what is new with taking a sample of the whole to come up with trends in statistics ?
Any concern should be whether 5% of 2 millions is a high enough sample (I would say so), and whether the samples is representative of the whole sample space (all the articles). The second I can't comment on as I do not know what those 100K articles are among other reasons.
KTC
-- Experience is a good school but the fees are high.
- Heinrich Heine
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
The editing sample is based on a random number generator selection of articles (namespace=0, page_is_redirect is false) in the September 8th dump of the page table. So it excludes articles created in the last weeks of September, but is otherwise a random sample of everything in article space.
As I alluded to before, though, there's a sort of [[survivorship bias]] in the fact that any articles deleted before September 8 are excluded from the sample.
And since you're not including redirects, there's also a (potentially large) bias against articles which were heavily edited in the past and then later turned into redirects.
Off the top of my head I can't think of any other significant problems.
I would expect the survivorship bias goes in the wrong direction though. Presumably pages that will be deleted, but haven't yet been, are more likely to be young. Hence the not-yet-deleted pages would seem to want to make recent edit counts higher. I can't think of any reason why survivorship effects would lead to a bump 6+ months ago.
And such biases would have no impact on the analysis of account creation, protections, or blocking. All of which also show drops.
-Robert
On 10/10/07, Anthony wikimail@inbox.org wrote:
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
The editing sample is based on a random number generator selection of articles (namespace=0, page_is_redirect is false) in the September 8th
dump
of the page table. So it excludes articles created in the last weeks of September, but is otherwise a random sample of everything in article
space.
As I alluded to before, though, there's a sort of [[survivorship bias]] in the fact that any articles deleted before September 8 are excluded from the sample.
And since you're not including redirects, there's also a (potentially large) bias against articles which were heavily edited in the past and then later turned into redirects.
Off the top of my head I can't think of any other significant problems.
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
On 10/10/2007, Robert Rohde rarohde@gmail.com wrote:
I would expect the survivorship bias goes in the wrong direction though. Presumably pages that will be deleted, but haven't yet been, are more likely to be young. Hence the not-yet-deleted pages would seem to want to make recent edit counts higher. I can't think of any reason why survivorship effects would lead to a bump 6+ months ago.
And such biases would have no impact on the analysis of account creation, protections, or blocking. All of which also show drops.
If the deletion rate has remained constant (as a proportion of articles), then you're absolutely right. If it's changed, then that could introduce a more complicated bias than just overestimating the recent edit counts.
Well if you look further down the analysis page, the rate of article deletion has also been declining. By contrast, article creation is flat or somewhat positive (depending on how long an interval one uses to estimate the trend).
-Robert
On 10/10/07, Thomas Dalton thomas.dalton@gmail.com wrote:
On 10/10/2007, Robert Rohde rarohde@gmail.com wrote:
I would expect the survivorship bias goes in the wrong direction though. Presumably pages that will be deleted, but haven't yet been, are more
likely
to be young. Hence the not-yet-deleted pages would seem to want to make recent edit counts higher. I can't think of any reason why survivorship effects would lead to a bump 6+ months ago.
And such biases would have no impact on the analysis of account
creation,
protections, or blocking. All of which also show drops.
If the deletion rate has remained constant (as a proportion of articles), then you're absolutely right. If it's changed, then that could introduce a more complicated bias than just overestimating the recent edit counts.
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
Well if you look further down the analysis page, the rate of article deletion has also been declining. By contrast, article creation is flat or somewhat positive (depending on how long an interval one uses to estimate the trend).
I'd imagine there must be a bump in a lot of these numbers at the time unregistered article creation was turned off. Have you seen anything on that one? (I don't remember when that happened, I'll look it up for you if you want).
On 10/10/2007, Anthony wikimail@inbox.org wrote:
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
Well if you look further down the analysis page, the rate of article deletion has also been declining. By contrast, article creation is flat or somewhat positive (depending on how long an interval one uses to estimate the trend).
I'd imagine there must be a bump in a lot of these numbers at the time unregistered article creation was turned off. Have you seen anything on that one? (I don't remember when that happened, I'll look it up for you if you want).
Roughly speaking, the beginning of 2006.
On 10/10/07, Anthony wikimail@inbox.org wrote:
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
Well if you look further down the analysis page, the rate of article deletion has also been declining. By contrast, article creation is flat or somewhat positive (depending on how long an interval one uses to estimate the trend).
I'd imagine there must be a bump in a lot of these numbers at the time unregistered article creation was turned off. Have you seen anything on that one? (I don't remember when that happened, I'll look it up for you if you want).
Here's a couple more events to consider: when was the captcha for unregistered edits which add an external link added? how about the captcha for new article creation?
It'd be interesting to make up a timeline of all the events we can think of which might possibly affect these numbers, and then attach them to the chart.
On 10/10/2007, Anthony wikimail@inbox.org wrote:
Here's a couple more events to consider: when was the captcha for unregistered edits which add an external link added? how about the captcha for new article creation?
It'd be interesting to make up a timeline of all the events we can think of which might possibly affect these numbers, and then attach them to the chart.
Semiprotection (could go either way); resolution of 550 (probably uptick in blocks *and* editing activity).
If you like thinking about "events", cascading protection led to an ~80% drop in the protection rate for images. (Though image protection is only a few percent of all protections).
-Robert
On 10/10/07, Andrew Gray shimgray@gmail.com wrote:
On 10/10/2007, Anthony wikimail@inbox.org wrote:
Here's a couple more events to consider: when was the captcha for unregistered edits which add an external link added? how about the captcha for new article creation?
It'd be interesting to make up a timeline of all the events we can think of which might possibly affect these numbers, and then attach them to the chart.
Semiprotection (could go either way); resolution of 550 (probably uptick in blocks *and* editing activity).
--
- Andrew Gray
andrew.gray@dunelm.org.uk
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
My comment on the reliability isn't suggesting that sampling is completely unreliable. But sampling with no comment on the selection method? That sounds like bullshit to me, and I'm not going to trust what is quite possibly an unreliable sample of whole. If you want to me to trust your stats, you need to provide more info than that.
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
If you like thinking about "events", cascading protection led to an ~80% drop in the protection rate for images. (Though image protection is only a few percent of all protections).
-Robert
On 10/10/07, Andrew Gray shimgray@gmail.com wrote:
On 10/10/2007, Anthony wikimail@inbox.org wrote:
Here's a couple more events to consider: when was the captcha for unregistered edits which add an external link added? how about the captcha for new article creation?
It'd be interesting to make up a timeline of all the events we can think of which might possibly affect these numbers, and then attach them to the chart.
Semiprotection (could go either way); resolution of 550 (probably uptick in blocks *and* editing activity).
--
- Andrew Gray
andrew.gray@dunelm.org.uk
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
Quit being rude and have a little faith.
I already explained it is a random sample higher up the thread.
-Robert
On 10/10/07, Steven Walling steven.walling@gmail.com wrote:
My comment on the reliability isn't suggesting that sampling is completely unreliable. But sampling with no comment on the selection method? That sounds like bullshit to me, and I'm not going to trust what is quite possibly an unreliable sample of whole. If you want to me to trust your stats, you need to provide more info than that.
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
If you like thinking about "events", cascading protection led to an ~80% drop in the protection rate for images. (Though image protection is
only
a few percent of all protections).
-Robert
On 10/10/07, Andrew Gray shimgray@gmail.com wrote:
On 10/10/2007, Anthony wikimail@inbox.org wrote:
Here's a couple more events to consider: when was the captcha for unregistered edits which add an external link added? how about the captcha for new article creation?
It'd be interesting to make up a timeline of all the events we can think of which might possibly affect these numbers, and then attach them to the chart.
Semiprotection (could go either way); resolution of 550 (probably uptick in blocks *and* editing activity).
--
- Andrew Gray
andrew.gray@dunelm.org.uk
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
Dismissal of your conclusions and evidence is not the same as rudeness. A little faith? Are you kidding me? Broad claims such as the ones you make are not something I take on faith, and in case you've forgotten, AGF doesn't apply to statistical data. You're the only person that I've seen to date suggest that involvement in en-wiki is waning.
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
Quit being rude and have a little faith.
I already explained it is a random sample higher up the thread.
-Robert
On 10/10/07, Steven Walling steven.walling@gmail.com wrote:
My comment on the reliability isn't suggesting that sampling is
completely
unreliable. But sampling with no comment on the selection method? That sounds like bullshit to me, and I'm not going to trust what is quite possibly an unreliable sample of whole. If you want to me to trust your stats, you need to provide more info than that.
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
If you like thinking about "events", cascading protection led to an
~80%
drop in the protection rate for images. (Though image protection is
only
a few percent of all protections).
-Robert
On 10/10/07, Andrew Gray shimgray@gmail.com wrote:
On 10/10/2007, Anthony wikimail@inbox.org wrote:
Here's a couple more events to consider: when was the captcha for unregistered edits which add an external link added? how about
the
captcha for new article creation?
It'd be interesting to make up a timeline of all the events we can think of which might possibly affect these numbers, and then
attach
them to the chart.
Semiprotection (could go either way); resolution of 550 (probably uptick in blocks *and* editing activity).
--
- Andrew Gray
andrew.gray@dunelm.org.uk
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
Yes, refering to my work as "bullshit" and an "insult [to] our intelligence" is unnecessarily rude.
Even if you disagree, I expect you to be civil about it.
If you dislike my analysis or observations, feel free to do your own. Ideally, Wikimedia would provide their own comprehensive statistics, but their analysis program hasn't been able to run on enwiki for about a year.
However, there is no sampling involved in analyzing Wikipedia's logged actions (i.e. every logged action was used), and these likewise show declines in account creation, article deletions, protections, blocking, etc. You may be entitled to your own interpretation, but you are not entitled to a different set of facts.
-Robert
On 10/10/07, Steven Walling steven.walling@gmail.com wrote:
Dismissal of your conclusions and evidence is not the same as rudeness. A little faith? Are you kidding me? Broad claims such as the ones you make are not something I take on faith, and in case you've forgotten, AGF doesn't apply to statistical data. You're the only person that I've seen to date suggest that involvement in en-wiki is waning.
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
Quit being rude and have a little faith.
I already explained it is a random sample higher up the thread.
-Robert
On 10/10/07, Steven Walling steven.walling@gmail.com wrote:
My comment on the reliability isn't suggesting that sampling is
completely
unreliable. But sampling with no comment on the selection method? That sounds like bullshit to me, and I'm not going to trust what is quite possibly an unreliable sample of whole. If you want to me to trust
your
stats, you need to provide more info than that.
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
If you like thinking about "events", cascading protection led to an
~80%
drop in the protection rate for images. (Though image protection is
only
a few percent of all protections).
-Robert
On 10/10/07, Andrew Gray shimgray@gmail.com wrote:
On 10/10/2007, Anthony wikimail@inbox.org wrote:
Here's a couple more events to consider: when was the captcha
for
unregistered edits which add an external link added? how about
the
captcha for new article creation?
It'd be interesting to make up a timeline of all the events we
can
think of which might possibly affect these numbers, and then
attach
them to the chart.
Semiprotection (could go either way); resolution of 550 (probably uptick in blocks *and* editing activity).
--
- Andrew Gray
andrew.gray@dunelm.org.uk
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
I wouldn't doubt that growth of the site is waning. How much more can an already-popular website expotentially grow? Think about it, Wikipedia is currently in the top 8(?) most popular websites on the Earth. There's not too many internet users that don't know already about Wikipedia's existance in some way. Unless theres a sudden burst of popularity in the media like the Essjay incident or the Edit to the Chris Benoit article that made headlines, I don't expect Wikipedia to actually increase in popularity anymore than it already has. Wikipedia had a meteoric rise from 2003 to 2006/07, maybe the level of popularity has just stabilized, it happens to all sites eventually.
Steven Walling steven.walling@gmail.com wrote:
Dismissal of your conclusions and evidence is not the same as rudeness. A little faith? Are you kidding me? Broad claims such as the ones you make are not something I take on faith, and in case you've forgotten, AGF doesn't apply to statistical data. You're the only person that I've seen to date suggest that involvement in en-wiki is waning.
--------------------------------- Tonight's top picks. What will you watch tonight? Preview the hottest shows on Yahoo! TV.
On 10/10/2007, Moe Epsilon moe_epsilon@yahoo.com wrote:
I wouldn't doubt that growth of the site is waning. How much more can an already-popular website expotentially grow? Think about it, Wikipedia is currently in the top 8(?) most popular websites on the Earth. There's not too many internet users that don't know already about Wikipedia's existance in some way. Unless theres a sudden burst of popularity in the media like the Essjay incident or the Edit to the Chris Benoit article that made headlines, I don't expect Wikipedia to actually increase in popularity anymore than it already has. Wikipedia had a meteoric rise from 2003 to 2006/07, maybe the level of popularity has just stabilized, it happens to all sites eventually.
Yeah, but I remember thinking that most of the steps along the way as well. Alexa ranking is strictly for amusement purposes. In any case, it's still going up:
http://www.alexa.com/data/details/traffic_details?site0=wikipedia.org&y=... http://www.alexa.com/data/details/traffic_details?site0=wikipedia.org&y=...
Here's the request and traffic graphs for the last year or so. This is across all of Wikimedia, not just en:wp:
http://tools.wikimedia.de/~leon/stats/ http://tools.wikimedia.de/~leon/stats/reqstats/reqstats-yearly.png http://tools.wikimedia.de/~leon/stats/trafstats/trafficstats-yearly.png
Those *may* be going flat, but I'd be surprised.
- d.
On 10/10/2007, David Gerard dgerard@gmail.com wrote:
On 10/10/2007, Moe Epsilon moe_epsilon@yahoo.com wrote:
I wouldn't doubt that growth of the site is waning. How much more can an already-popular website expotentially grow? Think about it, Wikipedia is currently in the top 8(?) most popular websites on the Earth. There's not too many internet users that don't know already about Wikipedia's existance in some way. Unless theres a sudden burst of popularity in the media like the Essjay incident or the Edit to the Chris Benoit article that made headlines, I don't expect Wikipedia to actually increase in popularity anymore than it already has. Wikipedia had a meteoric rise from 2003 to 2006/07, maybe the level of popularity has just stabilized, it happens to all sites eventually.
Yeah, but I remember thinking that most of the steps along the way as well. Alexa ranking is strictly for amusement purposes. In any case, it's still going up:
http://www.alexa.com/data/details/traffic_details?site0=wikipedia.org&y=... http://www.alexa.com/data/details/traffic_details?site0=wikipedia.org&y=...
Here's the request and traffic graphs for the last year or so. This is across all of Wikimedia, not just en:wp:
http://tools.wikimedia.de/~leon/stats/ http://tools.wikimedia.de/~leon/stats/reqstats/reqstats-yearly.png http://tools.wikimedia.de/~leon/stats/trafstats/trafficstats-yearly.png
Those *may* be going flat, but I'd be surprised.
- d.
Raw traffic isn't going to map with content creation. the 99% 1% rule. In the past we have been able to stay above 1% of those viewing editing but the percentage will likely change with new viewing audience. I suspect net regulars are the group with the greatest number of 1%ers in and I doubt we have much growth potential left among the net regulars.
Of people I know in my profession who more or less live on the net, probably all of them have looked at WP, but I don't know of more than 1 or 2 who are really active. I'd guess there is the potential for at least 5 times the number of active editors. The idea of combining reading and editing is still very new.
On 10/10/07, geni geniice@gmail.com wrote:
On 10/10/2007, David Gerard dgerard@gmail.com wrote:
On 10/10/2007, Moe Epsilon moe_epsilon@yahoo.com wrote:
I wouldn't doubt that growth of the site is waning. How much more can an already-popular website expotentially grow? Think about it, Wikipedia is currently in the top 8(?) most popular websites on the Earth. There's not too many internet users that don't know already about Wikipedia's existance in some way. Unless theres a sudden burst of popularity in the media like the Essjay incident or the Edit to the Chris Benoit article that made headlines, I don't expect Wikipedia to actually increase in popularity anymore than it already has. Wikipedia had a meteoric rise from 2003 to 2006/07, maybe the level of popularity has just stabilized, it happens to all sites eventually.
Yeah, but I remember thinking that most of the steps along the way as well. Alexa ranking is strictly for amusement purposes. In any case, it's still going up:
http://www.alexa.com/data/details/traffic_details?site0=wikipedia.org&y=... http://www.alexa.com/data/details/traffic_details?site0=wikipedia.org&y=...
Here's the request and traffic graphs for the last year or so. This is across all of Wikimedia, not just en:wp:
http://tools.wikimedia.de/~leon/stats/ http://tools.wikimedia.de/~leon/stats/reqstats/reqstats-yearly.png http://tools.wikimedia.de/~leon/stats/trafstats/trafficstats-yearly.png
Those *may* be going flat, but I'd be surprised.
- d.
Raw traffic isn't going to map with content creation. the 99% 1% rule. In the past we have been able to stay above 1% of those viewing editing but the percentage will likely change with new viewing audience. I suspect net regulars are the group with the greatest number of 1%ers in and I doubt we have much growth potential left among the net regulars.
-- geni
WikiEN-l mailing list WikiEN-l@lists.wikimedia.org To unsubscribe from this mailing list, visit: http://lists.wikimedia.org/mailman/listinfo/wikien-l
On 10/10/07, David Gerard dgerard@gmail.com wrote:
http://tools.wikimedia.de/~leon/stats/ http://tools.wikimedia.de/~leon/stats/reqstats/reqstats-yearly.png http://tools.wikimedia.de/~leon/stats/trafstats/trafficstats-yearly.png
For the purpose of our discussion those graphs are useless without interpolation to hide the big daily and weekly cycles. :(
Smoothed graphs are much nicer to look at: http://commons.wikimedia.org/wiki/Image:Eswiki-viewership-trend-2007.svg
:)
Too bad thats the ES Wikipedia graph ><
Gregory Maxwell gmaxwell@gmail.com wrote:
For the purpose of our discussion those graphs are useless without interpolation to hide the big daily and weekly cycles. :(
Smoothed graphs are much nicer to look at: http://commons.wikimedia.org/wiki/Image:Eswiki-viewership-trend-2007.svg
:)
--------------------------------- Be a better Heartthrob. Get better relationship answers from someone who knows. Yahoo! Answers - Check it out.
On 10/10/07, Moe Epsilon moe_epsilon@yahoo.com wrote:
Too bad thats the ES Wikipedia graph ><
ES wiki asks nicely.
On 11/10/2007, Gregory Maxwell gmaxwell@gmail.com wrote:
On 10/10/07, David Gerard dgerard@gmail.com wrote:
http://tools.wikimedia.de/~leon/stats/ http://tools.wikimedia.de/~leon/stats/reqstats/reqstats-yearly.png http://tools.wikimedia.de/~leon/stats/trafstats/trafficstats-yearly.png
For the purpose of our discussion those graphs are useless without interpolation to hide the big daily and weekly cycles. :(
Smoothed graphs are much nicer to look at: http://commons.wikimedia.org/wiki/Image:Eswiki-viewership-trend-2007.svg
Cor. We got some pretty good takeup in Mexico in the last couple of months...
(I love how prominent the academic terms are in that)
Steven Walling wrote:
Dismissal of your conclusions and evidence is not the same as rudeness.
Arguably true. However, your posts still struck me as shockingly and pointlessly rude.
Robert clearly did a lot of work that nobody, you obviously included, has yet done. That you may--or may not--have a way to improve the work is not a reason to accuse him of insulting us. It's a reason to offer constructive solutions to improve things.
That I'd have to explain that elsewhere is reasonable to me; sometimes we all get carried away when they see what they think is a problem. But here, group collaboration has built a world-class reference work from nothing in a few short years. I'm dumbfounded that anybody involved in the project could forget the power of friendly collaboration around modest beginnings.
William
On 10/10/07, Steven Walling steven.walling@gmail.com wrote:
Dismissal of your conclusions and evidence is not the same as rudeness. A little faith? Are you kidding me? Broad claims such as the ones you make are not something I take on faith, and in case you've forgotten, AGF doesn't apply to statistical data. You're the only person that I've seen to date suggest that involvement in en-wiki is waning.
A number of Wikipedians have pointed out the changes (starting around a year ago) that en-wiki has gone through:
Among others, see: http://www.andrewlih.com/blog/2007/09/10/two-million-english-wikipedia-artic... http://www.andrewlih.com/blog/2007/06/28/wikipedia-plateau/ http://ragesossscholar.blogspot.com/2007/03/watershed-in-history-of-wikipedi... http://original-research.blogspot.com/2007/03/is-wikipedia-approaching-barri...
I actually found Robert's conclusions much less surprising than he did (perhaps because he is just back from an extended break, while I and others have been watching this happen for a while).
-Sage
On 10/10/2007, Sage Ross ragesoss+wikipedia@gmail.com wrote:
A number of Wikipedians have pointed out the changes (starting around a year ago) that en-wiki has gone through:
Among others, see:
http://www.andrewlih.com/blog/2007/09/10/two-million-english-wikipedia-artic... http://www.andrewlih.com/blog/2007/06/28/wikipedia-plateau/
http://ragesossscholar.blogspot.com/2007/03/watershed-in-history-of-wikipedi...
http://original-research.blogspot.com/2007/03/is-wikipedia-approaching-barri...
I actually found Robert's conclusions much less surprising than he did (perhaps because he is just back from an extended break, while I and others have been watching this happen for a while).
-Sage
Personally, I would suggest that Wikipedia has indeed become more bureaucratic, and it will progress little further until a rethink of the core ideology is considered, particularly wrt. to how to derive/amend policy, core policy issues, handling bias or concepts of truth, dispute resolution and what to do when there isn't consensus (i.e. no consensus for the status quo, no consensus for proposed or active changes). The whole idea that Wikipedia acts by consensus is a sham. It's not a democracy of course either, it's not even anarchy, or specifically authority-driven (dictatorial). In individual cases it's whatever people can get away with. That's not a good concept of consensus (i.e. "what sticks is there by tacit agreement"); it ignores the fact that rational people will eventually give up rather than deal with bullies and morons.
Zoney
On 11/10/2007, Zoney zoney.ie@gmail.com wrote:
it ignores the fact that rational people will eventually give up rather than deal with bullies and morons.
Sorry, that's not entirely fair. There are also people who will not back down from their point-of-view because they truly believe they are in the right (indeed sometimes there are). Sometimes it's just persistence winning the day rather than bullying or such - though I still contend that there are those who are overly forceful in running roughshod in discussions over other contributors, even if not always breaking technical constraints such as "no personal attacks". Also, while it is offensive to call specific people morons, I think that we can accept that there are some on Wikipedia merely through there being so many contributors. And unfortunately, we have to deal with them. My point is that current decision-making mechanisms don't really allow for avoiding endless disruption caused by such individuals.
Zoney
On 10/11/07, Zoney zoney.ie@gmail.com wrote:
On 11/10/2007, Zoney zoney.ie@gmail.com wrote:
it ignores the fact that rational people will eventually give up rather than deal with bullies and morons.
Sorry, that's not entirely fair. There are also people who will not back down from their point-of-view because they truly believe they are in the right (indeed sometimes there are). Sometimes it's just persistence winning the day rather than bullying or such - though I still contend that there are those who are overly forceful in running roughshod in discussions over other contributors, even if not always breaking technical constraints such as "no personal attacks". Also, while it is offensive to call specific people morons, I think that we can accept that there are some on Wikipedia merely through there being so many contributors. And unfortunately, we have to deal with them. My point is that current decision-making mechanisms don't really allow for avoiding endless disruption caused by such individuals.
No offense taken. Good points though. Making personal attacks is always bad, while cleverly baiting otherwise decent people into making personal attacks (e.g. by behaving in the exact manner described by said personal attack) is generally tolerated.
—C.W.
On 10/10/2007, Sage Ross ragesoss+wikipedia@gmail.com wrote:
On 10/10/07, Steven Walling steven.walling@gmail.com wrote:
Dismissal of your conclusions and evidence is not the same as rudeness.
A
little faith? Are you kidding me? Broad claims such as the ones you make
are
not something I take on faith, and in case you've forgotten, AGF doesn't apply to statistical data. You're the only person that I've seen to date suggest that involvement in en-wiki is waning.
A number of Wikipedians have pointed out the changes (starting around a year ago) that en-wiki has gone through:
Among others, see:
http://www.andrewlih.com/blog/2007/09/10/two-million-english-wikipedia-artic... http://www.andrewlih.com/blog/2007/06/28/wikipedia-plateau/
http://ragesossscholar.blogspot.com/2007/03/watershed-in-history-of-wikipedi...
http://original-research.blogspot.com/2007/03/is-wikipedia-approaching-barri...
I actually found Robert's conclusions much less surprising than he did (perhaps because he is just back from an extended break, while I and others have been watching this happen for a while).
Presumably an encyclopedia is finite, there's a finite number of things that a person would expect to find in an encyclopedia, and each article can only be so big, before it gets broken up and/or merged.
In software, and in a sense the wikipedia is software, we find that the number of bugs in a piece of software exponentially declines over time, as people find and fix them. Arguably people writing the wikipedia are finding and fixing bugs in it- things that are missing or wrong. That would argue that the graph of size against person-hours would exponentially approach the omega point. However, meanwhile the number of contributors/viewers is exponentially growing (assuming that a certain percentage of viewers are contributors).
That argues that the Wikipedia is tending to an 'omega point' where it is 'perfect', where you've recorded everything that should be in an encyclopedia!
So when you take account of both effects, growth of new users/editors and exponential decay of bugs, so far as the amount of work done goes, you get a curve that looks a bit like a normal distribution curve- the same shape as the peak-oil curve and for similar reasons. It's possible that the downturn is that curve starting to come down. If it is starting to come down, perhaps it means, like the peak-oil curve means that the oil is half used up, that the English wikipedia is half written (obviously except for new discoveries and news)!
-Sage
On 10/12/07, Ian Woollard ian.woollard@gmail.com wrote:
Presumably an encyclopedia is finite, there's a finite number of things that a person would expect to find in an encyclopedia, and each article can only be so big, before it gets broken up and/or merged.
Well, I suppose an encyclopedia is limited in some way by the number of atoms in the universe, but for practical purposes the only real limit is whatever the writers impose on it.
I think there are well over a billion topics to write about. The theoretical limit hasn't been approached yet.
On 13/10/2007, Anthony wikimail@inbox.org wrote:
Well, I suppose an encyclopedia is limited in some way by the number of atoms in the universe, but for practical purposes the only real limit is whatever the writers impose on it.
I think there are well over a billion topics to write about. The theoretical limit hasn't been approached yet.
Right. But perhaps that's not the question. The question is are there a billion topics that are encyclopaedic? Presumably for something to be encyclopaedic it would have to be potentially interesting to a large number of people, not just people that happened to have physical contact with a particular street or school for example. A lot of a 'billion topics' would be only interesting to very few, so most of the article wouldn't be notable, and the article would (or should) be very short- and there would be less contributors to it.
Don't encyclopaedias emphasise generality by their very nature and isn't generality a summary? And isn't a summary inherently relatively short?
You know what? Ultimately, I have my suspicions, but I don't know. But if the graph does continue to come down, then I think we will begin to know.
Ian Woollard wrote:
Don't encyclopaedias emphasise generality by their very nature and isn't generality a summary? And isn't a summary inherently relatively short?
Only encyclopedias that are explicitly trying to be general encyclopedias, and that have a physical limit to how many pages they can pack in.
Specialist encyclopedias can be extremely detailed. You could have an "Encyclopedia of North American Wildflowers", for example, and it'd have really specific and detailed articles.
Encyclopedias that don't have a physical limit to how many pages they can pack in could well take the same level of detail that specialist encyclopedias have and apply it to everything. There's no compelling reason I can think of not to. Since there haven't been a whole lot of those until now, though, there aren't any examples one can point to and so people don't often think of the possibility.
On 13/10/2007, Bryan Derksen bryan.derksen@shaw.ca wrote:
Ian Woollard wrote:
Don't encyclopaedias emphasise generality by their very nature and isn't generality a summary? And isn't a summary inherently relatively short?
Only encyclopedias that are explicitly trying to be general encyclopedias, and that have a physical limit to how many pages they can pack in.
Specialist encyclopedias can be extremely detailed. You could have an "Encyclopedia of North American Wildflowers", for example, and it'd have really specific and detailed articles.
Sure. But how many specialist encyclopedias are there? A book is typically a megabyte of text, and the wikipedia is already more like a gigabyte; we could swallow quite a few specialist encyclopedias whole without noticeably getting any bigger. In fact we probably already have or are well on the way, some bird encyclopedias may well be a subset of the wikipedia (to pick a random example).
And there is notability, we do tend to restrict ourselves; a lot of popular culture rightly or wrongly is being stamped out with great malice right now. (My guess is wrongly, but I don't particularly care that much.)
Encyclopedias that don't have a physical limit to how many pages they
can pack in could well take the same level of detail that specialist encyclopedias have and apply it to everything. There's no compelling reason I can think of not to.
I agree there's no physical limit. Perhaps the limit is in the definition: 'it's an encyclopaedia dammit!'
Since there haven't been a whole lot of
those until now, though, there aren't any examples one can point to and so people don't often think of the possibility.
In which case it won't make much difference, because there will be few edits.
I guess ultimately, the wikipedia can be looked at a bit like an economy. Economies grow by innovation. We've been riding the 'anybody can edit the wikipedia encyclopaedia' innovation.
Innovations usually follow a bell-shape curve. We would need a new innovation to avoid falling back down again, although the area under the curve will be the wikipedia, so there's no loss per se in falling back in terms of updates, and we will still doubtless end up with the worlds best general encyclopaedia- albeit a bit bereft of popular culture :-)
On 10/12/07, Ian Woollard ian.woollard@gmail.com wrote:
On 13/10/2007, Anthony wikimail@inbox.org wrote:
Well, I suppose an encyclopedia is limited in some way by the number of atoms in the universe, but for practical purposes the only real limit is whatever the writers impose on it.
I think there are well over a billion topics to write about. The theoretical limit hasn't been approached yet.
Right. But perhaps that's not the question. The question is are there a billion topics that are encyclopaedic?
Depends what you mean by encyclopaedic. In my mind the term "encyclopedic" mainly refers to the form of an article. A topic is "encyclopedic" basically if it's in the form of an encyclopedia article title. So pretty much any noun is an encyclopedic topic. Might not be notable, mind you, but it's encyclopedic.
Presumably for something to be encyclopaedic it would have to be potentially interesting to a large number of people, not just people that happened to have physical contact with a particular street or school for example.
I'd call that notable, not encyclopedic, personally, but I'd venture a guess my idea of "a large number" is much smaller than yours. 100 people interested in a topic is probably enough by my mind. Certainly every publicly traded company in the world (or private one owned by at least 100 people, or with at least 100 customers), every book (or song, or periodical, or painting, or car model) which sold 100 copies (plus at least the main author of each such work), every school, every professional sports player, every candidate for public office including write-in candidates who received the necessary signatures to be a valid write-in candidate, every geographical location with at least 100 residents, every species, every software program with at least 100 downloads, every bridge crossed by 100 cars a week, etc.
That's only what's interesting enough, mind you. In the case of some of those topics, especially living people, we might not want to write about them (for privacy reasons, or simply because it'd attract too much vandalism, the latter of which might be fixable if a decent stable versions feature is implemented). Even throw out all living people though and I think we could still come up with a billion topics.
It'd be interesting to form a list. One argument from so called deletionists which I find most difficult to argue with is that we shouldn't have an article on a "less notable" topic until first having near-complete coverage of the "more notable" topics. But even if we went with that, there'd still be nearly no limit to the number of things we could write about, it'd just have to happen in a particular order.
Matthew Brown wrote:
However, I suspect the low-hanging fruit is getting closer to being exhausted. The topics that your average 15-25 year old computer nerd in an English speaking country is likely to know about.
If these statistics are actually the beginning of a new trend (I don't think that's a given), then I think that's probably a decent explanation, though somewhat oversimplified. I think the really low-hanging fruit that you describe was mostly exhausted long ago, and what's reaching the tipping point (think logarithmic decay) is the stuff that you don't know off the top of your head but you can easily research on the web.
But there are also limitations we're imposing on ourselves, especially with regard to companies and products. I've anonymously written several articles on companies and products and seen them deleted. Some others were tagged for deletion and probably would have been deleted had I not disputed the deletion tag. Think of the hullabaloo caused over http://en.wikipedia.org/w/index.php?title=Mzoli%27s_Meats&oldid=15851119.... I don't know if the spam tag that was added was a bit of trolling or POINTing or what, but had that article been written by an anonymous user I don't think it would have survived.
On 10/13/07, Anthony wikimail@inbox.org wrote:
Matthew Brown wrote:
However, I suspect the low-hanging fruit is getting closer to being exhausted. The topics that your average 15-25 year old computer nerd in an English speaking country is likely to know about.
If these statistics are actually the beginning of a new trend (I don't think that's a given), then I think that's probably a decent explanation, though somewhat oversimplified. I think the really low-hanging fruit that you describe was mostly exhausted long ago, and what's reaching the tipping point (think logarithmic decay) is the stuff that you don't know off the top of your head but you can easily research on the web.
Oh yeah, I meant to add this but I got sidetracked. I think another part of it is that a lot of the most popular articles are reaching or have reached the limit of what a wiki process can produce. In that sense the low-hanging fruit has become the hardest to reach. It's my opinion, and it's a controversial one, that a wiki can't produce a perfect article and at some point the amount of regression is going to equal the amount of progress. If there were some way to get people to stop wasting their time on articles at that stage and focus instead on the Mzoli's Meats of the world I think there'd be plenty of growth ahead.
Other people say the exact opposite though, that we should focus on improving [[Islam]] and forget about [[Mzoli's Meats]]. And that's probably what's going to happen for the near future.
On 10/13/07, Anthony wikimail@inbox.org wrote:
Oh yeah, I meant to add this but I got sidetracked. I think another part of it is that a lot of the most popular articles are reaching or have reached the limit of what a wiki process can produce. In that sense the low-hanging fruit has become the hardest to reach. It's my opinion, and it's a controversial one, that a wiki can't produce a perfect article and at some point the amount of regression is going to equal the amount of progress. If there were some way to get people to stop wasting their time on articles at that stage and focus instead on the Mzoli's Meats of the world I think there'd be plenty of growth ahead.
Other people say the exact opposite though, that we should focus on improving [[Islam]] and forget about [[Mzoli's Meats]]. And that's probably what's going to happen for the near future.
What sucks about that is that articles such as [[Islam]] probably have all the attention they need and then some. Perhaps not all the quality-focused encyclopedists they need, but such people are often driven off by the constant refighting of real-world battles.
What is lacking, IMO, are good articles on a lot of general but non-contentious topics. It's a lot easier to write an article on a very specialised topic than a general one. The specialized topic is likely to be well-bounded and of obvious and simple scope. Sources for such articles are generally easier to find (because they are so specific, they're easy to search for) and not contentious. It's a lot harder to find authoritative sources for general information, ironically, even though it's easy to find non-authoritative ones.
Working against original research is hard in a generalist article. It's really easy to write one from a personal perspective, but a lot harder to pin everything down.
-Matt
On 10/13/07, Matthew Brown morven@gmail.com wrote:
Working against original research is hard in a generalist article. It's really easy to write one from a personal perspective, but a lot harder to pin everything down.
To clarify: I'm using 'original research' in Wikipedia's sense here, not the actual logical one. The conclusions drawn in a general article are unlikely to be original. They are, however, likely to be based on the beliefs of the author and informed by the general consensus in the field, which can be awfully hard to pin down in good references.
-Matt
On 13/10/2007, Matthew Brown morven@gmail.com wrote:
What is lacking, IMO, are good articles on a lot of general but non-contentious topics. It's a lot easier to write an article on a very specialised topic than a general one. The specialized topic is likely to be well-bounded and of obvious and simple scope. Sources for such articles are generally easier to find (because they are so specific, they're easy to search for) and not contentious. It's a lot harder to find authoritative sources for general information, ironically, even though it's easy to find non-authoritative ones.
Indeed. I cannot even begin to figure out how I would have a shot at writing [[Law]]. Or [[colour]] (spelling disputes aside, it's noncontentious!). Or [[water]].
Matthew Brown wrote:
What is lacking, IMO, are good articles on a lot of general but non-contentious topics. It's a lot easier to write an article on a very specialised topic than a general one. The specialized topic is likely to be well-bounded and of obvious and simple scope. Sources for such articles are generally easier to find (because they are so specific, they're easy to search for) and not contentious. It's a lot harder to find authoritative sources for general information, ironically, even though it's easy to find non-authoritative ones.
Partly I think it's because this is a greater-than-Wikipedia problem. We're good at summarizing consensus knowledge, but "how to best provide a general overview of a field" is frequently something about which no consensus exists outside Wikipedia. There are usually a few major textbooks, survey articles, etc. (sometimes many more than a few), often with significantly different approaches and decisions about what should be covered at the top-level and how. I know in my field the most popular textbook is also one of the most resented and criticized!
In more specific articles we deal with this by basically summarizing all the major views/approaches/etc., but doing that in a top-level article defeats the purpose of making it a readable introduction. In a handful of cases we can split them into separate articles, like "the [x school] view on [y]", which is a little easier to find a consensus about.
-Mark
On 10/12/07, Anthony wikimail@inbox.org wrote:
I think there are well over a billion topics to write about. The theoretical limit hasn't been approached yet.
However, I suspect the low-hanging fruit is getting closer to being exhausted. The topics that your average 15-25 year old computer nerd in an English speaking country is likely to know about.
-Matt
Matthew Brown wrote:
On 10/12/07, Anthony wikimail@inbox.org wrote:
I think there are well over a billion topics to write about. The theoretical limit hasn't been approached yet.
However, I suspect the low-hanging fruit is getting closer to being exhausted. The topics that your average 15-25 year old computer nerd in an English speaking country is likely to know about.
A lot of them haven't been exhausted, they've been forbidden by various notability policies.
On 13/10/2007, Matthew Brown morven@gmail.com wrote:
On 10/12/07, Anthony wikimail@inbox.org wrote:
I think there are well over a billion topics to write about. The theoretical limit hasn't been approached yet.
However, I suspect the low-hanging fruit is getting closer to being exhausted. The topics that your average 15-25 year old computer nerd in an English speaking country is likely to know about.
A distinction I've suggested is that there are a near-infinity of possible topics, but only a finite few million that are topics more than one person actually wants to write ;-)
The low-hanging fruit is exactly it - we'll surge ahead, fill up the "comprehensible stuff" (Western, Anglospheric, contemporary, non-technical), then slow to a trickle of a few hundred a day on Moorish musicologists of the ninth century, minor asteroids, Yet Another Species Of Beetle, rural villages in Shenzen, etc. This won't be due to any lack of topics so much as due to a lack of people wanting to write about those topics.
So we need to figure out ways to channel our effort into expanding and developing what we have - the big field for new articles, which we may start mining soon or may start in four years time, will be breaking existing articles up into sections and spinning them off as daughter articles.
On 10/10/07, Robert Rohde rarohde@gmail.com wrote:
I would expect the survivorship bias goes in the wrong direction though. Presumably pages that will be deleted, but haven't yet been, are more likely to be young. Hence the not-yet-deleted pages would seem to want to make recent edit counts higher. I can't think of any reason why survivorship effects would lead to a bump 6+ months ago.
You're probably right about the effects, but I'm not really sure. The choice to exclude redirects also would probably have a similar effect, but again I can't be sure. And the breakdown of registered/unregistered and reverted/nonreverted would probably also be affected. The conclusion that there has been a decline in edits recently would almost definitely still apply, but the shape of the curve would be different, the timing of the decline might be different, and the mix of edits would probably be different.
4-6 months ago there was a lot of deletion and conversion to redirects surrounding BLP issues. For instance, Daniel Brandt's article, which had around 3000 edits, was converted into a redirect. But this article had no chance of being picked by the random selection, because it was turned into a redirect.
And such biases would have no impact on the analysis of account creation, protections, or blocking. All of which also show drops.
Yes, I think you're right that the activity rate (in the article space, at least) has been declining. At the same time the traffic apparently has been increasing (according to alexa, at least). What about the activity rate in the non-article space?
On 10/10/2007, Anthony wikimail@inbox.org wrote:
And since you're not including redirects, there's also a (potentially large) bias against articles which were heavily edited in the past and then later turned into redirects.
Yeah, this is an interesting one.
Firstly, we're ignoring a population of pages which have edit history but no current content - much the same issue as with deleted pages.
Secondly, though, we have a problem that doesn't occur from deleted pages - merges. When an article is redirected, it's often because it's been merged into another page; the text is copied over and a note left in the history. This is fair enough for our purposes, but for automated analysis like this it causes a glitch; the multiple edits over a long period which created that text aren't considered, and we end up perceiving the content as created at another page, at a much later date, in a single edit, by (probably) an unrelated user.
So how many merges are there? Difficult question. Juggling some numbers I'd guess about 0.5-1% of our total pages have at some point been merged...
On 10/10/2007, Steven Walling steven.walling@gmail.com wrote:
You're trying to make an accurate judgment based on 100k of articles from a 2 million article field? Don't insult our intelligence.
I hear science has invented this thing called "statistics", using these novel ideas called "samples".
Assuming our 100k were selected randomly, and that some attempt was made to consider the effects of articles which have been deleted since they were created... I can't see any data on the selection method, though.
I would be interested to see the results which have counterparts in our records - article creation rate, say - being compared, to see if any systematic discrepancies emerge. It won't prove the sample is perfect, but it will let us know if we have major weirdness.
On Wednesday 10 October 2007, Steven Walling wrote:
You're trying to make an accurate judgment based on 100k of articles from a 2 million article field? Don't insult our intelligence.
That might insult your intuitions, but not a statistician's intelligence -- not that I'm one. If the sampling is random and we can make some standard assumptions, a 100K article sample size would be *overkill*, be it a 2 million or 20 million population.
On October 30, 2006 -- the last day for the WP Report statistics - I believe the number of new accounts created was 7940 . If we look at the previous 10 months the average is 9092 with a std. dev. 608. Robert says he notices -25% decrease, so I'm going to assume that's nearly 1985 less account creations or 5955 account creations in his sample.
Now then, what's the chance that Robert's value (5955) was simply the result of a freakish variation from his sample of a population that was continuing as it was, to say nothing of continued massive growth? (This is what I'm inferring you to be claiming.) Impossibly small. See: http://en.wikipedia.org/wiki/Statistical_power
On 10/10/2007, Robert Rohde rarohde@gmail.com wrote:
rate of the Wikipedia community has been declining. This can be seen in the rate of editing articles (-17%), the rate of new account registration (-25%), blocks (-30%), protections (-30%), uploads (-10%), article deletions (-25%), etc.
Another quick addition to the lengthy list of footnotes to these statistics - the upload rate declining is pretty much what I'd expect, as we're getting better and better at pushing material onto Commons and more and more of our prolific uploaders know to go there as a first step. Note the Commons upload rate has really picked up.
The rate of short-lived dubious free image uploads is no doubt growing, but not enough to offset this.
(I can't believe this didn't occur to me earlier)