1
$\begingroup$

Would a bounty generate an answer for this question? Is there someone who is willing to work on it for points or is it generally unanswerable or what? Nick already told me it was very localized and that I do agree.

https://stats.stackexchange.com/questions/97247/new-systematic-quantile-normalization

$\endgroup$
3
  • 5
    $\begingroup$ Short answer is that we can't predict with certainty; what's certain is only that a bounty does draw attention to a question. (I think you are alluding to my comment on a now deleted Meta question; that's hard for most people to find, not that the original would add much here. But I am one of several Nicks around here.) $\endgroup$
    – Nick Cox
    May 12, 2014 at 13:11
  • 4
    $\begingroup$ What the -very interesting- question about the "New Systematic Quantile Normalization" is missing, is its mathematical representation. Right now, you are asking people to be able to fully understand your verbal descriptions, then translate them into mathematics and then work on them. That's too much to ask, except if you happen to find somebody already proficient in the mathematical representation of sound and music. $\endgroup$ May 21, 2014 at 23:25
  • $\begingroup$ +1 @Alecos -- a good summary of the basic issue $\endgroup$
    – Glen_b
    Jun 2, 2014 at 2:36

1 Answer 1

8
$\begingroup$

Bounties do help, but they're absolutely no guarantee.

Before contemplating a bounty you should endeavor to make your question as clear as possible. That will usually involve a reduction in jargon, and explaining as if readers are broadly knowledgeable of basic statistical ideas, but not necessarily familiar at all with your particular application. Your question should not just speak to experts on your problem, nor just to statistical experts, but later readers who may be relative statistical novices who are also unfamiliar with your problem (but may nevertheless find parallels to you problem if they can come to understand it, and so may benefit from the answer).

Without that, I don't think there's likely to be much to gain from a bounty. Getting a post to this point may require a fair bit of effort.

For example, in spite of having done my PhD thesis on a time series topic, I didn't even follow the entirety of your paragraph labelled "1." -- and I didn't feel any impulse to try to put in the effort required to decipher it any further. Without a bit more explanation/clarity, no bonus would be likely to induce me to do so.

Tell you what ... if you make the question nice and clear*, preferably with pictures -- I'll gladly put a bonus on it myself. You may have to nudge me with a comment, though, so I know you want me to look.


I don't think the question is sufficiently clear yet, but I'll improve my offer - if any two moderate-reputation readers (>500 rep, say) here think the question is either reasonably clear or think it nevertheless stands a good chance of an answer with a bonus in its present form (and let me know they think so), I will put up a bonus of at least 100 points.

If the question is later modified, and two such users then decide it counts as above (and let me know), then as long as it hasn't already had a bonus I will put up a 100 point bonus (at least) at that time, as long as I am still an active user.

[So there's a really good reason to try to improve the question - you just have to get it to a point where two people think it's worth trying a bonus on, and I'll supply the bonus.]


*(think ELI5, and you probably won't go too far wrong)

$\endgroup$
17
  • $\begingroup$ Wow thanks will do $\endgroup$
    – user11279
    May 13, 2014 at 11:41
  • $\begingroup$ Did the sentences and pictures i added help enough or does it still need more work? $\endgroup$
    – user11279
    May 13, 2014 at 18:08
  • 1
    $\begingroup$ It's certainly more attractive (and that has its benefits), but I suspect it won't make what you're doing, nor trying to get, that much more clear to the average statistician. Definitions or explanations of terms should probably be in words & algebra, while pictures would convey data and summarize relationships. It may simply be my own ignorance of what you're doing, but I still don't understand what this is asking - I still don't follow even item 1. I think my knowledge of stats is fairly broad, so it likely leaves out a lot of people who could potentially offer some help as it is now. $\endgroup$
    – Glen_b
    May 13, 2014 at 21:25
  • $\begingroup$ Thank you I will continue to work on it $\endgroup$
    – user11279
    May 13, 2014 at 21:30
  • $\begingroup$ stats.stackexchange.com/questions/97594/… $\endgroup$
    – user11279
    May 13, 2014 at 23:25
  • 2
    $\begingroup$ I get the feeling that we're talking mutually at cross-purposes here. I know what dimensionality reduction is for example, but that's kind of broad and unspecific as a topic. All the issue is entirely hidden in that last sentence of your new post. What the heck are we talking about there, in terms of the data? Statistics is about data, and models for data, and making inferences on the basis of those models. We don't seem to have remotely touched on your actual data, your models, or your desired inferences. It's a bit like asking an engineer "how do I build a structure?". $\endgroup$
    – Glen_b
    May 13, 2014 at 23:43
  • $\begingroup$ sorry i really am trying...my data is vocal recordings...the model for my data is sheet music...the inference for my data is correlations between the two sheets i create.. $\endgroup$
    – user11279
    May 13, 2014 at 23:48
  • $\begingroup$ No, sorry, data are not "recordings". Data, ultimately are numbers, possibly derived from recordings. "Sheet music" is not directly a model, either, though it's arguably a way of representing one. When you say "correlations" there, can you define the term algebraically, please? It doesn't seem to mean what a statistician would mean by the term (but then again we don't seem to be using many terms to mean the same things). What, exactly are you talking about when you say correlation? $\endgroup$
    – Glen_b
    May 13, 2014 at 23:52
  • $\begingroup$ I have a thought: can you find a post here on CV that's talking about the same kind of thing you're after? There are over 36 thousand posts here, perhaps one is discussing something like what you need. That way, at least there's a different explanation of the kind of thing you seek. On the other hand, if you can't find even one, then at least in looking for one, you might see the ways that statistical questions are framed, which might give you more of a sense of what you haven't yet conveyed. $\endgroup$
    – Glen_b
    May 13, 2014 at 23:58
  • $\begingroup$ no one else in cv has asked about sufficient dimension reduction i dont know that there is a term for the extrapolation of data into a new dimension $\endgroup$
    – user11279
    May 14, 2014 at 0:09
  • 1
    $\begingroup$ CV has literally thousands of questions relating to dimension reduction in some form or other (though not always using those words), but I don't know what the "sufficient" explicitly means in your expression, so it's hard to tell which kinds of thing it relates to. There are also many questions that relate to inferring information in various ways, which may relate to your needs on the "new dimension" thing ... but again, it's not clear what you need, exactly. $\endgroup$
    – Glen_b
    May 14, 2014 at 0:19
  • $\begingroup$ en.wikipedia.org/wiki/Sufficient_dimension_reduction ? $\endgroup$
    – user11279
    May 14, 2014 at 0:26
  • 1
    $\begingroup$ Ah, you mean 'sufficient' in the statistical sense. Okay, thanks, that helps a lot. There are probably well over a thousand questions on sufficient dimension reduction without that term necessarily having been used. Any sufficient statistic that's of smaller dimension than the data involves a sufficient dimension reduction ... and that's almost every question that relates to sufficient statistics. $\endgroup$
    – Glen_b
    May 14, 2014 at 0:28
  • $\begingroup$ To discuss sufficiency for some $\theta$, we must have some model (as here for example). As far as I can see we don't presently have anything approaching that. If there's a distributional model, there would be a way to start to talk about likelihood, for example. $\endgroup$
    – Glen_b
    May 14, 2014 at 0:34
  • 2
    $\begingroup$ Like Glen_b, I can't make sense of (1) in the question--and I have formal training in western music theory as well as a little bit of mathematical and statistical knowledge. Parts of that description are just too vague ("the data ... inhabit the same vector space but not yet in the same Cartesian space" is, according to my understanding of vector spaces and Cartesian coordinates, nonsensical) and others appear to be contrafactual ("Western music is a periodic sample expressing a frequency but rounded to a few Hz" is not really correct). There are comparable problems with (2) through (4). $\endgroup$
    – whuber Mod
    May 14, 2014 at 19:11

You must log in to answer this question.