Wikipedia:Pending changes/Request for Comment 2012/Discussion: Difference between revisions

Revision as of 17:30, 31 May 2012

Revision as of 11:07, 1 June 2012

Line 515:

::::One of the closers has indicated that the close [[User_talk:The_Blade_of_the_Northern_Lights#RfC|may come fairly soon]], and it's unlikely that the close will be: "The winner is Option [whatever]. Have a nice day." It will be easier to chart a course of action after we see what they have to say. I'm hoping some of the Option 1 guys will talk it out and come up with a course of action, and the same for some of the Option 2 guys. - Dank ([[User talk:Dank|push to talk]]) 13:40, 27 May 2012 (UTC)

:Kww's statistical analysis is, unfortunately, seriously misguided. I don't have the time and won't claim the expertise to do a full exposition of the [[convergent series]] principles involved, but his conclusion that support would eventual drop to the 50% level involves a basic statistical mistake. The point that a series like this converges to isn't determined simply by its concluding pattern, because as participation declines the "weight" os subsequent parts of the series declines. It's a trivial exercise (albeit one that makes many of those who aren't statistically adept suffer from blinding headaches and bleeding eyes) to construct a series of the form (A1-B1)+(A2-B2)+(A3-B3) . . . where, after a certain point, the B-terms are always greater than the A-terms, so that the sum of each additional pair is less than zero -- but the series converges (levels off) at positive number, never equaling or falling below zero. Several of the examples given in [[convergent series]] demonstrate this principle. There's no reason to believe that the outcome wouldn't stabilize at 60% or higher as participation fell off. [[User:Hullaballoo Wolfowitz|Hullaballoo Wolfowitz]] ([[User talk:Hullaballoo Wolfowitz|talk]]) 17:30, 31 May 2012 (UTC)

::You need to go reread the convergent series article. I agree that my 50% number is only an estimate, but the problem being solved isn't a convergent series problem at all. Those are sums of fractions that have a stated relationship to each other. This is a poll sampling problem, which has to do with the relative ratio of the population sampled to the population at large and discounting biasing effects.—[[User:Kww|Kww]]([[User talk:Kww|talk]]) 11:07, 1 June 2012 (UTC)

===Timetime===