In which I post a bunch of meaningless statistics

January 4, 2007 5:12 PM

In which I post a bunch of meaningless statistics in the hope of sparking a more meaningful discussion than this one.
posted by tkolar to MetaFilter-Related at 5:12 PM (145 comments total)

For the 36 months from January 2004 through December 2006

48043 questions were posted, an average of just under 44 per day

1439 of them were anonymous. The most any single user posted was 140.

6853 unique users (excluding anonymous) posted questions
2811 users posted 1-2 total questions
2764 users posted 3-10 total questions
1241 users posted 11-52 total questions
37 users posted 53-140 questions
posted by tkolar at 5:13 PM on January 4, 2007

Someone will inevitably ask for this, so I'll just cut to the chase.

Top 15 AskMe questions askers
(by number of questions asked between 1/2004 and 12/2006)
Questions User

76 benjh
76 geoff.
79 o2b
79 rorycberger
82 luriete
83 jdroth
86 twine42
93 mecran01
94 stupidsexyFlanders
97 Grod
105 dobbs
106 drezdn
106 scarabic
133 grumblebee
140 signal
posted by tkolar at 5:13 PM on January 4, 2007

On a more interesting and potentially useful note:

Average Questions Per Day and Average Answers Per Question

(by month from 1/2004 through 12/2006)

Month Questions/Day Answers/Question
2004/01 21 12
2004/02 17 11
2004/03 18 12
2004/04 20 12
2004/05 19 12
2004/06 23 14
2004/07 22 13
2004/08 23 13
2004/09 25 13
2004/10 23 12
2004/11 31 14
2004/12 35 16
2005/01 34 16
2005/02 38 16
2005/03 39 13
2005/04 39 13
2005/05 36 14
2005/06 40 13
2005/07 41 13
2005/08 46 13
2005/09 45 14
2005/10 49 15
2005/11 57 15
2005/12 54 16
2006/01 59 14
2006/02 58 14
2006/03 62 15
2006/04 58 14
2006/05 62 14
2006/06 64 14
2006/07 66 14
2006/08 71 14
2006/09 63 13
2006/10 71 14
2006/11 72 13
2006/12 62 14

posted by tkolar at 5:14 PM on January 4, 2007

Last but not least, a graph of the data.

(If you're as bored as I am today and you'd like access to the underlying XML data file I cooked up, drop me mail. It's about 3 meg and includes the username, time, and number of comments of every post to AskMe for the last three years.)
posted by tkolar at 5:15 PM on January 4, 2007

Well, stats are good, so hoopla! (At the risk of a derail) I'm curious about how you got them. Care to indulge my curiosity?
posted by stavrosthewonderchicken at 5:16 PM on January 4, 2007

This is all well and good, but who are the top 15 anonymous askers?
posted by cortex at 5:18 PM on January 4, 2007

Or perhaps, I'll post a link that works.
posted by tkolar at 5:18 PM on January 4, 2007

also, that graph is hosed at the moment
posted by cortex at 5:18 PM on January 4, 2007

er never mind
posted by cortex at 5:19 PM on January 4, 2007

Cortex wrote...

This is all well and good, but who are the top 15 anonymous askers?

Well I don't want to reveal too much, but keep your eye out for user 17564.
posted by tkolar at 5:21 PM on January 4, 2007

Pretty cool, tkolar. How are you getting this data? Got any more?
posted by iconomy at 5:27 PM on January 4, 2007

I'm curious about how you got them. Care to indulge my curiosity?

The AskMetafilter Archives by Date pages.
posted by tkolar at 5:29 PM on January 4, 2007

Interesting that the answers per question has remained relatively steady while the number of questions has risen over time.

Also, that I don't see a prima facie coorelation between the number of questions and number of answers.

We may be in question overload, but people are still responding to them. Whether they're actually being answered, though....
posted by dw at 5:34 PM on January 4, 2007

tkolar: "140 signal"

Apparently they were right: You can't stop the signal.
posted by Plutor at 5:36 PM on January 4, 2007

I have it on good authority that all of the anonymous questions were posted by signal.
posted by found missing at 5:37 PM on January 4, 2007

Got any more?

I've got the data I mentioned above, but I'm having a hard time extracting any more useful information from it. I'm open to suggestions.

I find three things of note about the data so far:

1) The number of questions per month on metafilter has tripled since 2003, and the trend shows no sign of slowing down.

2) The vast majority of questions are being posted by low-volume askers. For me this casts some doubt on whether limiting the frequency of posts by individuals will do much. But who knows, it's worth a try.

3) AskMetafilter has a heartbeat! (Or I screwed up the data collection somewhere) On a more practical note, perhaps queueing questions to even out the number of questions per day will buy AskMe some time before it is overwhelmed.
posted by tkolar at 5:38 PM on January 4, 2007

If signal ever gets a sockpuppet [NOT RACIST] I hope the username is 'noise'.
posted by stavrosthewonderchicken at 5:39 PM on January 4, 2007

tkolar, the time-series plot is really nice. Can you do the same for the entire record (2004-2006)? And, when were matt's changes in waiting periods applied?
posted by carmina at 5:40 PM on January 4, 2007

[Will] limiting the frequency of posts by individuals do much[?]. But who knows, it's worth a try.

Serious suggestion: someone who knows more of statistics than I do could come up with a half-decent answer to this question.
posted by matthewr at 5:40 PM on January 4, 2007

Ok, but how do the questions by month stack against the number of active users this month. Also, did you not hear when Matt said to give this a rest until there have been a few months of data? Lastly, that thread is still open.
posted by Eideteker at 5:43 PM on January 4, 2007

The vast majority of questions are being posted by low-volume askers.

I would imagine so, there are so many users! Maybe it's worth to see how many users post at regular intervals? Know what I mean?
posted by carmina at 5:45 PM on January 4, 2007

[Will] limiting the frequency of posts by individuals do much[?]. But who knows, it's worth a try.

Serious suggestion: someone who knows more of statistics than I do could come up with a half-decent answer to this question.

Unfortunately I don't think statistics alone can tackle that. It's more of a systems question, where you would think (hope) that dampening one of the inputs would cause a disproportionate change in the overall system.
posted by tkolar at 5:46 PM on January 4, 2007

Finally, a useful, non-confrontational MeTa post. Thanks for the info. tkolar.
posted by SeizeTheDay at 5:53 PM on January 4, 2007

Two things strike me:
1. The number of answers per question is relatively unchanged, which suggests that there is no problem with the increased volume of questions.
2. The vast majority of users post very infrequently, so bumping the 1-week to 2-weeks is only going to annoy some people, and not for very good reason.
posted by knave at 5:54 PM on January 4, 2007

carmina wrote...
Can you do the same for the entire record (2004-2006)?

Yes, here you go.
posted by tkolar at 5:54 PM on January 4, 2007 [1 favorite]

thanks, oh mi god.
posted by carmina at 5:55 PM on January 4, 2007

well, clearly. One measure is to have people post questions relative to their answering ratio. There.
posted by carmina at 5:57 PM on January 4, 2007

Do you have data that could show the increased number of users along with the increased number of questions? Make the # of users a single line that follows the questions trend, if possible. Thanks.
posted by SeizeTheDay at 5:57 PM on January 4, 2007

140 signal
140 questions in 26 months means damn near one every single week for three years. I still don't see how anyone could have that many questions that are important enough to post here, yet difficult enough to not be easily solved with equivalent or less effort by any normally intelligent person.

It's quite possible that this is just me, of course.
posted by dg at 5:59 PM on January 4, 2007

Yeah, yeah, that should be 36 months, not 26.
posted by dg at 6:00 PM on January 4, 2007

How many questions were posted within less than 2 weeks of that person's last question?
posted by cillit bang at 6:00 PM on January 4, 2007

BTW, I'll be the first to admit that I was wrong if the real reason why questions are scrolling off the front page so quickly is the rapid rise of new users. The thought had occurred to me before, but never stuck in my brain. Makes sense, though I'm pretty sure shutting off new users is not an option Matt will take, since he now lives off of this site.
posted by SeizeTheDay at 6:01 PM on January 4, 2007

SeizeTheDay, are you advocating to close down signups? Because you know, it won't happen.
posted by carmina at 6:01 PM on January 4, 2007

tkolar, I know we are jumping on you, but cillit bang has an idea: can you plot the number of unique users as a function of the minimum posting interval (in weeks, but more than the waiting period) for each user?

I know, I know. But, but, we will die wondering.
posted by carmina at 6:04 PM on January 4, 2007

tkolar, this is all great but I think one major missing datapoint is the total users as a third line on the posts/day and answers/post graphs. I would bet the new user numbers correlate to the posts/day line, and if so, yeah, it's mostly new users that are adding to the question load and not too many questions from existing users.
posted by mathowie (staff) at 6:07 PM on January 4, 2007

Unfortunately I don't think statistics alone can tackle that. It's more of a systems question

I thought it would be possible to construct some kind of model for the number of questions asked per user per time period. I'd guess the number of questions asked by a user is a function of the amount of time since they joined up (scared newbies less likely to ask questions), the number of questions they've previously asked (past well-answered questions encourage more), and their activity on the site (measured by comments total) — and doubtless many other factors, which ought to be mostly measurable.

Given an individual function, you could construct an aggregate function and then look at the effect of imposing a two-week constraint.

But it's entirely possible that what I've said is nonsense. For a start, I know nothing of systems theory (but I'm studying econometrics at the moment). It could be an interesting project though.
posted by matthewr at 6:07 PM on January 4, 2007

he vast majority of users post very infrequently, so bumping the 1-week to 2-weeks is only going to annoy some people, ...
By that logic, the number of people annoyed is a minute minority, with no down side for the vast majority, yet an upside for the not insignificant number who agree that many people ask way too many questions. Sounds like the right decision has been made.
posted by dg at 6:08 PM on January 4, 2007

mathowie: it's mostly new users that are adding to the question load

I agree that new users is the factor likely to correlate most strongly with the questions/day count.

Doesn't that suggest, however, that the whole question-per-two-weeks thing is unlikely to have much long-term effect?
posted by matthewr at 6:10 PM on January 4, 2007

Eidetaker wrote...
Ok, but how do the questions by month stack against the number of active users this month.
SeizeTheDay wrote...
Do you have data that could show the increased number of users along with the increased number of questions?

I thought about these but I'm having a hard time coming up with a way to know the "number of active users" in any given month. It's just data that I don't have.

The total number of accounts is fairly useless as there's no way to tell how many are still in use.

(and I'm not going to run a script against every single user checking last login. I'm just gathering some easy stats, not trying to crash the server)

Also, did you not hear when Matt said to give this a rest until there have been a few months of data?

Actually I did miss that, probably because I bailed out of the other thread when I realized that people were have a joyous shouting match where the lack of underlying information was something of a benefit.

In any case, this thread wasn't meant to be a discussion of the two week rule. It's mostly meant to provide some solid information about what's been going on in AskMe over the last three years.
posted by tkolar at 6:12 PM on January 4, 2007

No, dg, because to me it seems that the people who are going to get burned are probably people who don't post too often, but happen to have two questions in a two week period. Myself I've posted maybe 10 questions per year, and I would definitely hit the two-week wall a few times because of the sporadic nature of my questions.

The people who are posting every single week are clearly very few in number, and really, what's the problem? We're seeing no reduction in the quantity of answers (although we can't measure quality), so it seems like there was no real problem to solve here.
posted by knave at 6:13 PM on January 4, 2007

Other desired plots:
- upper and lower bounds (uh, quartiles?) on answers per day (what was the average number of answers for the top 25% most-answered questions each day? Bottom 25%?)
- total answers per day (which the near-constant average answers per question value implies would correlate strongly with questions-per-day)
posted by cortex at 6:14 PM on January 4, 2007

Doesn't that suggest, however, that the whole question-per-two-weeks thing is unlikely to have much long-term effect?

Yeah, that's what I was saying. Looking at the data, if I could add registered users/day to it and get a good fit to the posts/day line, that would suggest posts per 1 or 2 weeks wouldn't have much difference.

The most important takeaway is the average answers/question is relatively stable, even as "stuff scrolls off the front page way too fast" these days.
posted by mathowie (staff) at 6:15 PM on January 4, 2007

yeah, to back up cortex, I kind of wish I could know the median and standard deviations for answers/post and posts/day to know how much variability there is.
posted by mathowie (staff) at 6:15 PM on January 4, 2007

and I'm not going to run a script against every single user checking last login. I'm just gathering some easy stats, not trying to crash the server

Run it against every 100th user number, get a rough estimate of monthly milestones from there. Low impact, good enough for gov't work.
posted by cortex at 6:16 PM on January 4, 2007

Yeah, "stuff scrolls off the front page way too fast" is a meaningless complaint if the increasing number of users is the main driver of the increasing number of questions — especially given answers/question is constant.
posted by matthewr at 6:17 PM on January 4, 2007

*coughcough* db dump pester pester *coughcough*
posted by cortex at 6:18 PM on January 4, 2007

Hmm, interesting. If you look at the 2006 graph, you can see the effect of the outage in September. No one could post questions on the 13th, so the 14th had almost double the average number of posts.
posted by Partial Law at 6:18 PM on January 4, 2007

godd job tkolar

That still leaves the issue of how best to keep questions from rolling off the page too quickly. You can make the page bigger, you can restrict the question size and force "more inside" (although didn't that already happen) and you can break them down into a few catagories, among other things. If you go with catagories I suggest keeping it to a minimum so people hit them all in their browsing. That will maximize the number of people who see the question, and for the obscure questions this is really important.
posted by caddis at 6:20 PM on January 4, 2007

good enough for gov't work

Wild derail here, but I looked up that phrase (not having heard it before) and found this mildly interesting rant, which includes the counter-intuitive origin of the phrase.
posted by matthewr at 6:20 PM on January 4, 2007

Thinking about this more, I bet the only reason the answers/post has been relatively stable as the posts have grown is due to the increase in users offering answers.

I gave jessamyn some interesting data a few months ago, but I seem to recall the number of users of mefi that have left any answer was pretty high, like around 28,000, so most every user registered here had helped out at one time or another.
posted by mathowie (staff) at 6:21 PM on January 4, 2007

That still leaves the issue of how best to keep questions from rolling off the page too quickly.

I don't think it does. If, as Matt believes, new users are the driving force behind the increasing number of questions, and given the fact the answers per question are constant, the same number of people see a question posted now as saw questions posted in the past.
posted by matthewr at 6:23 PM on January 4, 2007

Weird/neat, matthewr. I'd love to see a better source on that origin than an angry editorial pep talk, though.

*goes poking*
posted by cortex at 6:25 PM on January 4, 2007

the same number of people see a question posted now as saw questions posted in the past.

That's a dangerous leap to conclusion, matthewr. I don't know if it's possible to calculate whether the same number of people saw a question now as before unless we were able to calculate pageviews per question from when the site first started until now. And even then, that doesn't account for whether the sheer volume of questions causes the page to scroll to fast that people don't bother clicking to the second or third frontpage.
posted by SeizeTheDay at 6:27 PM on January 4, 2007

mathowie writes 'The most important takeaway is the average answers/question is relatively stable, even as "stuff scrolls off the front page way too fast" these days.'

It struck me in the last thread that a lot of the people who complain about the too-fast scrolling read AskMe as entertainment, as well as treating it as a resource that you're obliged to contribute to as well as use. Not that there's anything wrong with reading Ask for kicks, but it doesn't matter if there's too much for people to keep up with in that sense, as long as the answers keep coming.
posted by jack_mo at 6:30 PM on January 4, 2007

I object to those numbers on the grounds of observer bias. Flawed! Flawed, I say!
posted by blue_beetle at 6:31 PM on January 4, 2007

I thought about these but I'm having a hard time coming up with a way to know the "number of active users" in any given month. It's just data that I don't have.

You've got the unique number of posters asking questions. If you could add the same thing for answers then you'd have a measure of active users, for askme at least. They might not be currently active but have shown some sign of life at some stage weeding out the total lurkers or non-activated accounts. But I'm guessing you're not going to be able to do that (comments are less accessible right?).

Thirding the wish for quartiles and measures of variability and adding to the overall awe of what has been produced.
posted by shelleycat at 6:36 PM on January 4, 2007

That's a dangerous leap to conclusion, matthewr.

Is it? This graph shows that answers per question is remarkably constant, despite huge increases in users and questions.

When I Ask Metafilter, all I care about is the quality of the answers. This is hard to quantify, so (assuming we're not somehow getting stupider over the years) the quantity of answers is not a bad proxy variable for this.

For some reason, people don't look at the number of answers per question, but instead consider the amount of time their question is on the front page. There's no good reason for this at all — what matters is the number of answers (and this hasn't changed).

Of course, if there are more questions then each question scrolls off the front page faster. But if, as people seem to agree, the number of questions is highly correlated with the number of users, there's no reason to think that your question is seen by fewer people merely because its on the front page for less time - surely the increased number of users compensates for that.
posted by matthewr at 6:36 PM on January 4, 2007 [1 favorite]

Unless you have access to data that suggests that pageviews cause answers, you have a "correlation does not equal causation" problem with your argument, matthewr, which is why your conclusion is dangerous.
posted by SeizeTheDay at 6:45 PM on January 4, 2007

Unless you have access to data that suggests that pageviews cause answers

I don't have data, but it seems like a reasonable assumption. Why would pageviews not increase answers more-or-less proportionately?

Are new users (who are presumably the prime cause of increased pageviews) less likely to answer questions? That, to me, seems improbable, and sounds like simple nostalgia - especially given the trend of new users who join up just for AskMe.

Of course, I don't deny that there are various correlation/causation issues here, but I'm advancing a theory that sounds pretty reasonable to me, not writing for Econometrica.
posted by matthewr at 6:53 PM on January 4, 2007

tkolar, this is really fascianting, thanks.
posted by jessamyn (staff) at 6:54 PM on January 4, 2007

Yeah, "stuff scrolls off the front page way too fast" is a meaningless complaint if the increasing number of users is the main driver of the increasing number of questions — especially given answers/question is constant.

Not true. At a certain point, you might get the same number of questions, but the speed of the site will be so rapid it will lose any sense of being a coherent community, which is what makes AskMetafilter successful where other similar sites have failed miserably.

Now, whether this is avoidable is another question entirely. I suspect that it won't be possible to completely curb the influx, but any reasonable actions to do so have my backing.

I already avoid AskMe a lot of the time because it moves far too quickly for me to reasonably keep track of what is going on. Some people don't care about this, others do. On the other hand, I've been reading metafilter long enough to recall the hand-wringing over the blue getting too busy. A similar move to "one post a day" rather than allowing people to post two, three, or four times a day mitigated this somwhat. But, more than that, opening a dialogue about keeping things controlled made people think more carefully about whether or not the post needed to be made.

Similarly, people posing questions on Askme might run into the two-week barrier (if only once) and think to themselves, "you know, I didn't realize this was an issue. Perhaps I'll think more carefully about what questions I ask in the future". It certainly doesn't hurt.
posted by The God Complex at 6:55 PM on January 4, 2007

Not true. At a certain point, you might get the same number of questions, but the speed of the site will be so rapid it will lose any sense of being a coherent community, which is what makes AskMetafilter successful where other similar sites have failed miserably.

But that isn't necessarily true, or true to the degree that you may be supposing. There is a meta-layer of coherent community to AskMe, sure, and to AskMe-as-part-of-mefi, but on an answer-by-answer basis I'd wager that a lot of users (and especially the newer folks) are connecting more to individual questions (and to a small subset of similar-minded answerers) than they are to AskMe as a gestalt.

AskMe can continue to function very well without the insularity of, say, Metatalk—the Cheers of Metafilter, no doubt—so long as their is a sense of standards and general community. The five dollar entrance fee helps there, as does the as-needed moderation.
posted by cortex at 7:00 PM on January 4, 2007

I need to run out for a bit, but I'll see if I can generate a few more graphs before the evening is through.


mathowie writes...
I kind of wish I could know the median and standard deviations for answers/post and posts/day to know how much variability there is.

Unfortunately I forgot how to do either of those a long time ago, and I'm afraid if I just pull in packages I'll misapply them and end up giving bogus answers. If someone more statistically savvy than me wants to take a shot at it, drop me an email and we'll work out a way to get the data to you in a useful format.
posted by tkolar at 7:06 PM on January 4, 2007

At a certain point, you might get the same number of questions, but the speed of the site will be so rapid it will lose any sense of being a coherent community, which is what makes AskMetafilter successful where other similar sites have failed miserably.

I've always thought that the community cohesion of AskMe (which I agree is our USP) is mostly exogenous — generated by MetaFilter and the 'community policing' of MetaTalk. Is there really a sense of community in AskMe itself? Discussion is limited (by moderation) to answers to the question in hand — I don't think you could 'get to know people' purely in AskMe.

As the 2004-06 graph shows, questions have trebled in three years. Has there really been much loss of community cohesion in that time? I don't see AskMe as being in decline at all — for example, popular blogs (like LifeHacker) hold AskMe in high regard and repost questions weekly; we get mentioned in newspapers; and as MetaTalk threads show, regularly win various awards/accolades/etc.
posted by matthewr at 7:07 PM on January 4, 2007

On (not) preview, what cortex said.
posted by matthewr at 7:09 PM on January 4, 2007

BRONZE!!

Suck it, haters.
posted by scarabic at 7:20 PM on January 4, 2007

tkolar, I could help. Email me.
posted by carmina at 7:24 PM on January 4, 2007

Thanks tkolar, this aligns with the premise that euphorb presented yesterday - more members means more questions, but also more answerers, and hence steady number of responses per query.
posted by Mister_A at 7:24 PM on January 4, 2007

At a certain point, you might get the same number of questions, but the speed of the site will be so rapid it will lose any sense of being a coherent community

I agree. Maybe to approach it a little differently, the greater volume of questions makes it that much harder for any given user to keep up with all of the questions/discussions going on. If community = most of the users reading most of the stuff on the site, then I do think we are losing something as the scrolling gets faster.
posted by Mid at 7:25 PM on January 4, 2007

I've always thought that the community cohesion of AskMe (which I agree is our USP) is mostly exogenous — generated by MetaFilter and the 'community policing' of MetaTalk. Is there really a sense of community in AskMe itself? Discussion is limited (by moderation) to answers to the question in hand — I don't think you could 'get to know people' purely in AskMe.

I often wonder about this, and didn't mean to suggest that I thought it was an absolute that a sense of community would be lost. Do note, however, that there is a reasonably large contingency of users on AskMetafilter that exclusively use AskMetafilter and not the rest of the site--at least as an active participant. I've noticed this quite often, as I have a habit of perusing posting/commenting histories when I come across interesting answers/questions (or posts, when it comes to the blue or grey).

There's a definite sub-sect in AskMetafilter, but I confess to not being certain how much community exists there. In fact, since the "community" of the blue/grey is what I find so inviting, I often find AskMe cold and alienating--but these people were initially attracted to the site for a different reason than I was. In fact, owing to that, I'm not even sure if my opinion of AskMe is all that relevant. For the first while that it existed, I tried (and failed) to stop myself from making jokey comments, and had several deleted (including a couple great ones, curse the Gods!). After that, I realized why I wasn't meshing with the site well and now I mostly avoid it, for the better of the community.

But I also spend far too much time thinking about these things, so...

PS - Thanks for the data, tkolar. Very neat.
posted by The God Complex at 7:29 PM on January 4, 2007

Median and standard deviation are easy and Excel will give perfectly adequate values. Median, in particular, is just the middle number of a set and can be worked out just by counting in from each end if you have the patience. I wouldn't use Excel for anything more complicated (mean and sum are generally OK) and I'm not completely sure the best way to work out a quartile, but these simple descriptive statistics are pretty difficult to screw up too much.
posted by shelleycat at 7:49 PM on January 4, 2007

Hm. So the quantity of answers to question has remained the same, and measure the quality of answers is pretty much impossible. There are a ton of questions being posted, but its not like there aren't possible solutions:

* where are the per category rss feeds? If I'm a huge mac nerd, and I just want to subscribe to computer questions, I don't see how I can add this to my rss reader.

* you could make the front page personalized, where you configure the categories of questions you see by default, and the rest would be hidden unless you explicitly showed them.

* you could open up tagging to be real tagging, where anyone can tag stuff. Yes, you would get junk, but you would also get better tag results over time. I don't think matt is on board with that, though.

* allowing viewing/sorting by user, page views, # of answers, etc. Basic filtering stuff.

Just some ideas...personally I just browse when I have time, and never come here when I'm busy. I don't see the problem with a lot of volume if people are still gettin quality answers.
posted by rsanheim at 7:52 PM on January 4, 2007 [1 favorite]

the greater volume of questions makes it that much harder for any given user to keep up with all of the questions/discussions going on. If community = most of the users reading most of the stuff on the site, then I do think we are losing something as the scrolling gets faster.

This would seem to suggest perhaps a limit of one question per month or two to ensure the volume of questions can't get any bigger as membership grows, or that membership should be curtailed.

I'm glad to hear that the number of answers/question is stable, as it points to the site continuing to scale well (assuming 14 answers today are about as good as 14 answers were three years ago). Maybe the whole "scrolling off the page too fast" is a metric of perception that the data doesn't support being an issue, but it feels like it should be.

Alternately, I could up the questions per page to 100 up from 50, to ensure that an entire day fits on the front page, but that would tend to promote more questions, wouldn't it?
posted by mathowie (staff) at 7:52 PM on January 4, 2007 [1 favorite]

rsanheim, you should subscribe to the "mac" tag feed then. I subscribe to the "tivo" feed and try and answer all questions that come down on it.
posted by mathowie (staff) at 7:54 PM on January 4, 2007

If community = most of the users reading most of the stuff on the site, then I do think we are losing something as the scrolling gets faster.

Granted, but it's something that we can't avoid losing without radically altering (and constraining) use of AskMe, which isn't going to happen.

Note also that it's only a small core of mefites who can be said to actually read most of the stuff on the blue—what we'd be losing (or have, really, already lost) doesn't even exist on the blue.

In the other direction, there may actually be value in a high volume of questions: with such an abundance of queries, avid (or bored) users are more likely to encounter questions on which they have legitimate knowledge, and thus may be less likely to quell their appetite with blind stabs in questions they're less qualified to answer.
posted by cortex at 7:54 PM on January 4, 2007 [1 favorite]

Oh, finally, I should point out that if the site grows too much, the moderation will be stretched thin (Jessamyn already does an incredible job for the volume of questions she has to deal with). There's always a danger when moderation has to be expanded to meet site demand--that is to say, a community site is only as good as its moderators, and finding able fill-ins that will maintain a sense of cohesion with the current site is not an easy task. See, for example, how long it took Matt to even add a second "voice of authority" to MeFi proper.
posted by The God Complex at 7:58 PM on January 4, 2007

Maybe the whole "scrolling off the page too fast" is a metric of perception that the data doesn't support being an issue, but it feels like it should be.
Not everything can be supported (or otherwise) by data. Some things just are.
posted by dg at 8:10 PM on January 4, 2007

assuming we're not somehow getting stupider over the years

I don't know about this. I think it's more likely that the chances that a question will be definitively answered after 15 attempts is pretty good.

It occurs to me that it might make jessamyn's job easier to queue questions until, say, 9 AM EST the next day, the same way that Anon AskMe questions are queued. That way jess could get up, have a swim and a cup of tea, wander in lazily at 10 AM, identify and remove all the "My disgustingly fat girlfriend wants an abortion - can I hit her?" questions, and be ready for an early lunch.
posted by ikkyu2 at 8:18 PM on January 4, 2007

QUEUEING BAD!

*chews scenery*
posted by cortex at 8:19 PM on January 4, 2007

with such an abundance of queries, avid (or bored) users are more likely to encounter questions on which they have legitimate knowledge, and thus may be less likely to quell their appetite with blind stabs in questions they're less qualified to answer.

This is it exactly. A steady number of answers per question does not mean that we are getting the same quality of answer or the same depth of participation. I suspect there is no way to measure quality/depth (though the average number of words in each comment might be a proxy), but I do think we have to be losing something with the volume.
posted by Mid at 8:21 PM on January 4, 2007

By "this is it exactly," I actually meant the opposite, I now realize.
posted by Mid at 8:24 PM on January 4, 2007

What'd you use to make the graphs?
posted by smackfu at 8:24 PM on January 4, 2007

Excellent job, tkolar.

The answers per question metric maybe isn't all that great. People join, people want to contribute, so they answer questions. That doesn't mean the answers are any good. I really do think there is something to the speed of the scrolling on the front page. I asked a question yesterday which seemed like the perfect AskMe question: it was about a movie, about an obscure part of the movie, about information not readily available, but which the users here might have some insight into. It got six responses, only two of which were from someone who appears to have seen the movie. I'm glad that those who answered did, but, and I recognize that this is only anecdote, I really feel like this is a question AskMe would have been able to answer if more people had been able to see the question.
posted by OmieWise at 8:25 PM on January 4, 2007

So, like, I don't really have much invested in this either way, but it'd be really cool if the preview page (after a question was composed and previewed) were more proactive, and came back with a whole table of stuff related to the question through some magical AI -- related tags, similar questions already asked, the ones that had 'best answers' marked, those answers, the Asker's shoe size, and so on, and then, in big yellow letters "HAVING REVIEWED PAST QUESTIONS, ARE YOU SURE YOU STILL WANT TO ASK THIS, PUNK?" or something to that effect.

That'd be a help, right there. I'm serious.
posted by stavrosthewonderchicken at 8:26 PM on January 4, 2007 [2 favorites]

That way jess could get up, have a swim and a cup of tea, wander in lazily at 10 AM, identify and remove all the "My disgustingly fat girlfriend wants an abortion - can I hit her?" questions, and be ready for an early lunch.

I don't even wake up until 10 or 11, and I don't wan tto face a question queue first thing in the am.

I do think there is a cohesiveness to the AskMe community that is different from other parts of the site, and having a manageable amount of questions helps that. My feeling, based on not much, is that if a question gets at last a few early answers, that people reading it via my comments or the tag feeds will see it or keep an eye on it making sure it doesn't go abandoned.

When mathowie ran some numbers for me I was suprised that the number of unanswered questions remained more or less constant over time and I'd still like to encourage use of the STUMPED tag for people who are unhappy to have gotten no answers or none that really answered the question. The unanswered page is suprisingly empty, though it is based on an answer count of zero.

My only other suggestion would be a week or so waiting period for your first question on AskMe, the same as on the Blue. I find this discussion more interesting than the other -- similar in so many ways and yet not -- AskMe thread in MeTa just a few doors down.
posted by jessamyn (staff) at 8:31 PM on January 4, 2007

Interesting idea, stavros. Let the post-preview page do the search they should have done in the first place. Would probably require a bit of hackery to come up with a good way to isolate likely keywords.

Strip common tokens from the post text, do a non-phrasal search against the least-common tokens (or just a weighted search against all tokens favoring rarity for matches) and present a short list of "hey check these out first, why doncha" links for the poster to (hopefully) peruse before proceeding.

Half the questions I've gone to ask in the last few months have been abandoned after I found answers in the archive by some googlry.
posted by cortex at 8:34 PM on January 4, 2007

It appears that I was right.
posted by Afroblanco at 8:37 PM on January 4, 2007

(does an irritating little dance)
posted by Afroblanco at 8:39 PM on January 4, 2007

Yeah, that's what I was saying. Looking at the data, if I could add registered users/day to it and get a good fit to the posts/day line, that would suggest posts per 1 or 2 weeks wouldn't have much difference.

The most important takeaway is the average answers/question is relatively stable, even as "stuff scrolls off the front page way too fast" these days.

I was fucking right and I didn't need no stinkin' data. Remember the other time I was right? Yeah. So bring back the image tag, too. Cause I'm right.
posted by dame at 8:41 PM on January 4, 2007

Kudos to tkolar.

Another option of limiting questions instead of the one/two week limit is basing the allowable frequency on the user's historical answer to question ratio.

Something like: At your present ratio of answers to questions, you have 03 days left before your next question. Of course, the obvious problem with this is that it encourages more, but not necessarily better answers.

Oh, and I totally agree with what stavros was saying.
posted by philomathoholic at 8:42 PM on January 4, 2007

dame, I love you.

(In a totally platonic way, of course)

(although you're wrong about the image tag)
posted by Afroblanco at 8:55 PM on January 4, 2007

Are there any readily available statistics on AskMeFi pageviews? I'm curious what kind of correlation there would be between increased daily questions and increased viewership. Possibly, with the knowledge that answers fly on the green so fast, users are also frequenting the page more often. If anything, the decreasing lack of downtime between questions might interest some users to become more involved.
posted by jmd82 at 9:04 PM on January 4, 2007

I'm by no means well-known 'round these here parts, but I think I should speak up for that part of the "silent majority" which I feel I belong. I'm one of those people who hangs out on the green, and pops over to the blue and grey when there's a lull in questions. :)

Do things move quickly? Yep, but it doesn't bother me - I tend not to comment unless I think I've actually got something to contribute. I do read nearly every post to the green, however, and find that even during my hectic day it's very easy to keep up. So far, there have been a vanishingly small number of subjects sought for which someone has already posted, and no useful answer was contributed, and most of those are TravelFilter stuff. AskMe is an invaluable resource for my 100% travel job!

Also, I belong to the larger majority of infrequent posters. I've had a few questions kicking around, but nothing worthy of a post, and the 2-week limit doesn't come close to touching me.

I do want to note that I'm absolutely opposed to limiting post frequency based on answer frequency - that would be the fastest way to raising the noise floor of AskMe to unusable levels.

My $0.80 (Chicago tolls are cheap!).
posted by TheNewWazoo at 9:13 PM on January 4, 2007

Alternately, I could up the questions per page to 100 up from 50, to ensure that an entire day fits on the front page, but that would tend to promote more questions, wouldn't it?

Why not just always show the last 24 hours of questions? Hitting "older questions" goes back 24 hours. Hitting "newer questions" goes forward 24 hours. You could make the RSS feed for the last 24 hours as well.

I'm not sure if it promotes more questions, but it seems possible that it would also promote more answers.
posted by sequential at 9:25 PM on January 4, 2007

Great stuff, tkolar, thanks.

The "But questions scroll too fast to get answered!" complaint has always simply accepted as given that questions are increasingly suffering from faster scrolling off the front page. Your info at least tries to measure that, and suggests the complaint is a bit hollow.

The God Complex: I already avoid AskMe a lot of the time because it moves far too quickly for me to reasonably keep track of what is going on.

What do you mean by "keep track of what is going on"? You seem to have an expectation that members are reading every question, or need to read every question, as if "community" necessarily suffers when most members *don't* read every thread. But how long has it been since the majority of members here have been reading all of *any* part of the site? How many members click through to every discussion in the blue on an average day? I'd bet it's not a majority (by a long shot), which means it's certainly not a requirement for the "community" here that we all read every thread. Hasn't been for years, and the site still seems to be doing fine.

Why "avoid AskMe" when you can just choose to look at the questions that interest you and skip the ones that don't? Is it somehow hard for you to discriminate between questions like that? No snark intended; I'm honestly curious why you use such an all-or-nothing approach.

SeizeTheDay: whether the sheer volume of questions causes the page to scroll to fast that people don't bother clicking to the second or third frontpage.

Why don't they bother? Members who claim to care about AskMe but find themselves stopped by the awful barrier of clicking to the 2nd or 3rd page of questions are being silly at best. I dunno, it seems the idea that questions scroll off the front page too fast to get answered has gotten far more traction than it deserves, and tkolar's added information solidifies that.
posted by mediareport at 9:48 PM on January 4, 2007

Makes sense, though I'm pretty sure shutting off new users is not an option Matt will take, since he now lives off of this site.

I hear a fair amount about closing signups - which obviously isn't going to happen - but what about raising the cost of signups? $10 memberships would raise more revenue for the site and also slow down the tide of new members some.
posted by joshuaconner at 9:52 PM on January 4, 2007

The most important takeaway is the average answers/question is relatively stable, even as "stuff scrolls off the front page way too fast" these days.

HOLD ON!!!

I want everyone to re-examine this assumption and what it means. The data doesnt support this, really.... Best to illustrate by example:

The category "human relations" generally gets the most comments - everyone likes to chime in on other's problems. Lets assume the number of human relations questions GO UP not just absolutely, but as a proportion of all questions. That is, Ask.Metafilter has gotten not better but "chattier"

Lets also assume the average for ALL OTHER types of questions (travel, music, ID questions, etc.) go DOWN!

Result: the average answers/question stays the same and yet that doesnt mean what you think it means...

This concludes your Statistics 101 lecture.
posted by vacapinta at 9:52 PM on January 4, 2007

Lets assume the number of human relations questions GO UP not just absolutely, but as a proportion of all questions.

I see the general point you're making, but the specific example is a bad one. I think a better assumption is that chattier questions have actually been policed *down* as a proportion of all questions over the past two years.
posted by mediareport at 10:05 PM on January 4, 2007

*cue tkolar on another stats run*
posted by mediareport at 10:05 PM on January 4, 2007

I did the analysis two weeks ago and came to the same conclusion. We discussed it yesterday as well.
posted by euphorb at 10:30 PM on January 4, 2007

I don't necessarily have the right answers for you, mediareport. It's not a process of logical reasoning that brought me to that feeling--it's more of a feeling. If on Metafilter every day there were seventy-five posts with three-hundred comments in each resulting discussion, I'd probably stop reading that too. It would feel impersonal to me, much the same way some people bemoan their favorite mom and pop shop being put out of business by a big box store. The attendant rhythms are lost.

But, as I made clear before, my opinion about Ask Metafilter might not be entirely relevant. It has its own community (some of it shared, some of it not). Honestly, part of my problem is that many of the questions put far too much of the information inside the thread, so much so that you don't even know what the question is about. It's hard to do some selective self-filtering that way.

One final thought. You mentioned that it's not necessary for people to read every part of the site every day. I agree. But, from my perspective, I appreciate at least being able to leisurely look at the thirty or forty blue posts in a day and then deciding what interests me enough to read through (for one reason or another). I do the same in metatalk. I sort of do the same in AskMe (though I didn't for some time). I'm not necessarily suggesting we're at critical mass now--I'm simply suggesting that if measures aren't taken, the efficacy of moderation will suffer, and the sheer volume of questions will make the site ungainly and impesonal: I have no interest in trying to filter my way through 150 questions on a given day. Will it end up at that point? I have no idea. It might. And I think AskMe would be most effective if any given user could, in a reasonable timeframe, scan over the questions--if not for interest, than at least to see if they could help out (which makes the site all the better, in my mind).

I could be way off base. This is just how I've felt about it for some time.

(note that, as I said earlier, I think Jessamyn is doing a terrific job there; I also don't think the moderating would be as consistently good if it was the purview of more people).
posted by The God Complex at 11:07 PM on January 4, 2007

In the other direction, there may actually be value in a high volume of questions: with such an abundance of queries, avid (or bored) users are more likely to encounter questions on which they have legitimate knowledge, and thus may be less likely to quell their appetite with blind stabs in questions they're less qualified to answer. posted by Cortex

Hear, hear!
posted by IndigoRain at 11:48 PM on January 4, 2007

smackfu wrote...
What'd you use to make the graphs?

Excel. (if anyone knows of a good OS X graph creation program I'm all ears)
posted by tkolar at 1:14 AM on January 5, 2007

Okay, I got distracted from creating new graphs based on the old data and went and collected a bunch of new data instead. I'll mix the data together tomorrow and see if I can answer some of the more direct questions that have been asked.

In the meantime:

There are roughly 28770 registered Metafilter users

Of those, 14770 have made one or more comments (answering questions is counted as a comment)

3716 users have posted more than 100 comments total.

Graphs of new users per day are here.

I'm admittedly a little less certain of this data. The user pages are a lot more finicky to extract data from.

It's way past my bedtime. G'night.
posted by tkolar at 1:22 AM on January 5, 2007

Now if you could only find a way to measure the quality of the answers.
posted by SteveInMaine at 1:37 AM on January 5, 2007

Hmmm. it'd be interesting to see who had the highest ratio of "answers marked best' to "total answers given".
posted by pjern at 3:13 AM on January 5, 2007

This is some fascinating stuff. I'm not really convinced that just because the answers/question ratio is the same that each question is actually getting an appropriate number of answers. Seems to me (in my experiene of reading a fair whack of the askme questions that get posted every day) that there's a not small number of questions that get way more than 15(ish) answers, which would tend to skew the average a bit higher. And upon further review of the thread, people asking for stdev and median are already way ahead of me :)
posted by antifuse at 4:01 AM on January 5, 2007

But do the numbers prove they aren't getting answered? Is there a such thing as an appropriate number of answers? Are people worried about volume just convinced that there is a Platonic Better Answer out there that was lost because the question wasn't on the front page long enough? I asked a question that got two answers, but one of them was all I needed. Not to mention, from what I recall, there were more chatty, everyone can answer questions in the beginning.
posted by dame at 6:06 AM on January 5, 2007

The only valid response to the complaint that questions drop off the front page too fast is "read AskMe more often".
posted by grateful at 6:41 AM on January 5, 2007

The God Complex nails it. x2.
posted by Mid at 6:45 AM on January 5, 2007

My only other suggestion would be a week or so waiting period for your first question on AskMe, the same as on the Blue.

We already have that.
posted by mathowie (staff) at 7:20 AM on January 5, 2007

How about a "recent questions" sidebar thing showing the next 25 older questions that are on the 2nd page?
posted by amberglow at 7:24 AM on January 5, 2007

dame nails it. x2.

1. The number of answers per question is relatively unchanged, which suggests that there is no problem with the increased volume of questions.
2. The vast majority of users post very infrequently, so bumping the 1-week to 2-weeks is only going to annoy some people, and not for very good reason.

Yup.

Members who claim to care about AskMe but find themselves stopped by the awful barrier of clicking to the 2nd or 3rd page of questions are being silly at best.

Double yup.
posted by languagehat at 7:29 AM on January 5, 2007

I have no interest in trying to filter my way through 150 questions on a given day.

Why do you have to? Why not just "filter your way" through the first 40 or 50 you encounter until you get tired/bored/burned out, and then move on?

And I think AskMe would be most effective if any given user could, in a reasonable timeframe, scan over the questions--if not for interest, than at least to see if they could help out (which makes the site all the better, in my mind).

Ok, fair enough. But you *can* scan over the questions in a reasonable timeframe now, if by "reasonable timeframe" you mean "under ten minutes." Go ahead, time yourself: Can you scan 10 new questions in one minute? Of course you can, and that's 100 questions in ten minutes. (I just slowly scanned the newest 10 questions in 30 seconds, actually, so in ten minutes I could probably scan 200 questions).

If someone can't spare ten minutes once every two or three days to scan a few pages of AskMe questions, they really have no business complaining about the growth of the site.

That said, I do think you're over-estimating the importance to "the community" of everyone having a meaningful encounter with every question that gets posted. MeFi has been a large, segmented set of mini-*communities* for years now. Most folks participate in the subset of threads they're interested in and move on, and yet somehow whatever communal sensibility there is manages to get passed on - mainly, I think, by keeping every post listed on one main front page.
posted by mediareport at 7:34 AM on January 5, 2007

...and came back with a whole table of stuff related to the question through some magical AI -- related tags, similar questions already asked, the ones that had 'best answers' marked, those answers, the Asker's shoe size, and so on, and then, in big yellow letters "HAVING REVIEWED PAST QUESTIONS, ARE YOU SURE YOU STILL WANT TO ASK THIS, PUNK?" or something to that effect.

I like this idea. I have used a couple of tech support websites that, when you previewed your support request before submitting it, came back with links to the knowledge base/FAQs based on a search for words in your request. It'd be useful on AskMe since there seems to be an increase in duplicated questions recently.
posted by EndsOfInvention at 7:39 AM on January 5, 2007

"It's not a process of logical reasoning that brought me to that feeling--it's more of a feeling."

Are you hooked on this feeling? Because unless you've got some data, it can never really be more than a feeling.

And as to the "quality of answers" question, would the proportion of answers marked best be a decent metric? I realize that they haven't been going too long, the best answers, and that they're used inconsistently, but I'm curious about their trends. Aside from that, better answers weigh more, so we can sort threads by weight ;)
posted by klangklangston at 8:02 AM on January 5, 2007

This is a side thing, but did anybody notice that this guy, CitrusFreak12, has shot up the Contribution Index by posting 65 comments in his first week? He's at #4 today, and he was at #3 yesterday. I almost emailed him to tell him he should keep commenting at least twenty times a day to stay up there, but I thought he might take that challenge seriously.
posted by koeselitz at 8:20 AM on January 5, 2007 [1 favorite]

I picture him huddled into a cubicle somewhere, wolfing down oranges and commenting furiously.
posted by koeselitz at 8:21 AM on January 5, 2007

whole oranges are too slow—it's Tang-boosted grapefruit-/orange-juice cocktail by the gallon
posted by cortex at 8:35 AM on January 5, 2007

I'm also impressed that I guess we managed to register 11 CitrusFreaks before CitrusFreak12 got here.
posted by koeselitz at 8:37 AM on January 5, 2007

Late to the party, but psyched about the graphs.

Way up-thread, Partial Law notes that the 2006 graph shows the September outage, and the next day, with its big spike in the number of questions. What's really neat about that is that you DO see the corresponding dive in the number of answers per question. The one that you DON'T see on tkolar's lovely second graph. (Which, if you didn't click it before, and you're still reading, click it click it! It makes the more-questions-but-steady-answers trend crystal clear. Euphorb was first with the identical message, but if looks count, tkolar's drives it home.)

Slightly expanded conclusion: When more questions don't result from more users, you get fewer answers per question. But, when more questions do result from more users, answers per question is steady.
posted by daisyace at 9:23 AM on January 5, 2007

I'm heartily sick of the one and two-week limits. It's basically resulted in me asking only every couple of months, because when I think about asking a question, I hold off because what if I really need to ask one tomorrow? So questions hang about in my head for months until I need the answer, pronto.

Still, I suppose that's what it was designed to do. It just feels a bit of a gyp, giving answers then being unable to ask.
posted by bonaldi at 10:11 AM on January 5, 2007

bonaldi, everyone should be like you tho--you give more than you get, which is wonderful.
posted by amberglow at 11:26 AM on January 5, 2007

For those of you curious about what the distribution of questions/answer looks like, here it is for my small dataset.

Days= 12 (spread over three years)
Total questions= 8921
Total answers= 639
Mean A/Q= 14.0 +/- 13.0
Median A/Q= 10

So the few questions with a large number of answers do skew the mean A/Q slightly.
posted by euphorb at 11:52 AM on January 5, 2007

A look at the "unanswered questions" tab is fairly revealing - 9 unanswered questions since 1 December 2006. Not a lot IMO.
posted by Mister_A at 12:17 PM on January 5, 2007

How many users posted questions within 2 weeks of one another?

I've only included data from 2005 and 2006, since January 2005 appears to be when the 1 week limitation went into effect. (To give you an idea, 92 of signal's 140 questions were asked in 2004)

6620 users asked questions in 2005 and 2006.
39488 questions were asked in those two years combined.

Of the 39488 questions, 7796 were asked by a user within two weeks of asking a previous question.

Of the 6620 users, 2136 users asked at least one question within two weeks of another.

-----------

Top fifteen sub-two-week posters by number of sub-two-week posts:
28 signal
30 Kickstart70
32 WCityMike
34 Raybun
34 curiousleo
34 k8t
34 krisjohn
35 sdis
36 snsranch
36 zek
38 grumblebee
38 xmutex
42 sirion
48 zardoz
50 rolypolyman
posted by tkolar at 12:54 PM on January 5, 2007

How many users signed up just to post a question?

452 users have more questions than answers to their names. The worst is that bastard Anonymous, who has asked over 1400 questions and never answers anything.

225 users have asked questions but never left answers.
posted by tkolar at 1:22 PM on January 5, 2007

Can I make a vague request for some Music analysis? I've found myself wondering about:
- number of posters vs. number of commenters
- proportionate size of the intersection of posters and commenters
- percentage of poster/commenters in Music whose on-site activity is significantly within Music
posted by cortex at 1:29 PM on January 5, 2007

tkolar: "225 users have asked questions but never left answers."

Names. This information is no good without names.

And home telephone numbers.
posted by koeselitz at 4:32 PM on January 5, 2007

This concludes your Statistics 101 lecture.

The concept of a "median" should really be in Stat 101. The Median won't go up as a small subset of threads get chattier.
posted by scarabic at 6:11 PM on January 5, 2007

Of the 6620 users, 2136 users asked at least one question within two weeks of another.

Wow, that's a lot. Perhaps Matt's idea of tightening that will actually help reduce questions after all.

bonaldi: I'm heartily sick of the one and two-week limits. It's basically resulted in me asking only every couple of months, because when I think about asking a question, I hold off because what if I really need to ask one tomorrow?

Oh come on. You've got so much goodwill built up here, bonaldi, you could *easily* get someone to ask an emergency question for you via email or just by posting it to MeTa. "Heartily sick"? Methinks you're being over-dramatic. There will NEVER be a time when you can't get an emergency AskMe answered, so you should stop withholding your questions and just ask the questions you want when you want.
posted by mediareport at 6:17 PM on January 5, 2007

BTW, I think it sucks that all attempted solutions to the problem of underserved threads have been about reducing the number of questions instead of increasing the number of answers.

Hm... questions aren't getting enough answers? They're scrolling off the page too fast? Let's just eliminate a bunch of them so they don't get answered AT ALL (as they can't be asked). Not the most productive approach.

How about linking answering frequency to the ability to ask more questions? Or "best answer" frequency? A "best answer" buys you a free question with a 1-week vesting cliff. I'm sure it's been suggested.

I guess anything can be abused. You could beg some asker you know to "best answer" you since it costs them nothing. But the 1 week cliff would cut down on that kind of desperation abuse.

In the current setup, I guess the point is that "abuse" means getting another account and putting another $5 in the till. Nevermind. Cancel everything I said. I understand now.

Couldn't we have just made the sockpuppet fee $10 instead of cutting frequency to once every two weeks?
posted by scarabic at 6:25 PM on January 5, 2007

I kind of wish I could know the median and standard deviations for answers/post and posts/day to know how much variability there is.

Okay, the first of those is ready to go. Mean, median, and standard deviations for answers per question. Once again I've done two, one with the outliers removed so the graph will be a bit more readable.

(in this case removing outliers means that anything with a standard deviation > 40 had it's standard deviation reduced to 40)

I've also made the graphs much much bigger.


In case you were wondering, 10/2/2006 was "purification.org" day.
posted by tkolar at 10:49 PM on January 5, 2007

I kind of wish I could know the median and standard deviations for answers/post and posts/day to know how much variability there is.

Here's the second part, posts/day data calculated month by month.

The graph is here, and the data is below.


(month, mean, median, standard deviation)

01/2004 21.48 21.00 6.40
02/2004 17.88 17.00 6.02
03/2004 18.61 19.00 6.94
04/2004 20.50 20.00 4.85
05/2004 19.71 18.00 7.42
06/2004 23.13 23.00 7.16
07/2004 22.29 22.00 6.96
08/2004 23.16 24.00 7.93
09/2004 25.57 25.00 8.99
10/2004 23.42 22.00 7.52
11/2004 31.17 31.00 9.41
12/2004 35.68 34.00 12.89
01/2005 34.87 37.00 8.63
02/2005 38.50 39.50 9.48
03/2005 39.45 42.00 11.78
04/2005 39.63 41.50 10.46
05/2005 36.55 35.00 11.45
06/2005 40.87 42.00 10.82
07/2005 41.84 45.00 14.76
08/2005 46.52 50.00 13.76
09/2005 45.53 44.00 13.37
10/2005 49.10 49.00 12.38
11/2005 57.27 57.50 12.40
12/2005 54.58 61.00 17.06
01/2006 59.10 59.00 13.48
02/2006 58.00 57.00 14.15
03/2006 62.39 64.00 13.82
04/2006 58.87 59.50 14.95
05/2006 62.26 62.00 12.97
06/2006 64.40 68.50 15.34
07/2006 66.45 67.00 18.25
08/2006 71.81 77.00 17.74
09/2006 63.17 64.00 22.45
10/2006 71.19 75.00 16.48
11/2006 72.20 76.50 18.50
12/2006 62.87 65.00 19.23
posted by tkolar at 11:10 PM on January 5, 2007

cortex wrote...
Can I make a vague request for some Music analysis?

That's probably not going to happen this time around. Sorry.

I'm still working working on some of the other questions though.
posted by tkolar at 11:28 PM on January 5, 2007

standard deviation as a measure of variability is only interesting when adjusted for changing means. So the apparent increase in variability in posts/day through time is actually a decrease.
posted by Rumple at 12:42 PM on January 6, 2007

Why not make the limit 2 questions per 30 days? If somebody is afraid of wasting their one question, in case a second more pressing question comes, they can still use that second question tomorrow. However, if they use two in two days, then they will then need to wait another 29 days before asking another question.
posted by stovenator at 10:07 PM on January 6, 2007

So I was thinking.... if I hypothetically were to have all of the comment data, what kind of information would be interesting? Here's what I've got so far

1) Active users each day and month (top ten most consistent users)
2) Average words per post overall and per user (top 10 posts and users also)
3) average links per post per user (top 10 posts and users also)
4) What percentage of users make 90% of the posts?

Anything else?
posted by tkolar at 7:18 PM on January 7, 2007

Oh, and by "post" I mean posts, comments, questions, and answers
posted by tkolar at 7:19 PM on January 7, 2007

Oh yeah, I should also do an average comments per hour on a metatalk post, to pinpoint the exact moment it's not worth posting to it anymore because everyone else has moved on.
posted by tkolar at 8:10 PM on January 7, 2007

And while I have the floor here, I'd like to take a moment to thank whomever added these users for really screwing up the screen scraper.

12292

13286

13287

13288

13289

13299

13300

13306

13307

13308

13309

13310

13311

13315

13318

posted by tkolar at 8:19 PM on January 7, 2007

tkolar, that was in the 5k contest days, when people were fucking around. The bug has been patched, but I'll remove those users.
posted by mathowie (staff) at 11:29 PM on January 7, 2007 [1 favorite]

Wow, someone actually does read this far down in these threads. Thanks for removing the whacked out users.

Just in case you don't have this from your own tools, I did finally plot the number of active users per day. I don't have it on a graph vs. the other data yet, though.

Oddly enough, the average users per day appeared to go *up* during the 2004 holidays.
posted by tkolar at 12:54 AM on January 8, 2007

Whups, I forgot to mention that the above data is only for AskMe.
posted by tkolar at 9:09 AM on January 8, 2007

Koeselitz, Cortex, you've both made my day (albeit a little late).
posted by CitrusFreak12 at 1:01 PM on January 9, 2007