3 ms·

Didn't read the link, but could a mod please change this title, which is obviously false? No correlation would mean that if I bought a random headphone that co

Didn't read the link, but could a mod please change this title, which is obviously false? No correlation would mean that if I bought a random headphone that cost $2 (they exist, you can go to ali express right now and put in a maximum of $2 in a headphone search), and a random headphone that cost $500, then if you had to make a bet about which one would come closer to reproducing the bass of a song with a heavy bass, you would be betting even money. It would be a toss-up whether the $2 or the $500 came closer to producing that bass. Because there is no correlation. Here is an example of correct usage of "no correlation": there is no correlation between a headphone's price and the md5 checksum of its SKU. EDIT: I skimmed the paper. A better title (for HN) would be "No correlation between frequency response and price quartile in 283 headphones".

navbaker 9y ago
If you read the link, you'll find it's the exact title of the journal article.
- logicallee 9y ago
  It is still obviously false. It is a sensational title and should be changed so that it is true. (It's like saying, "Scientists find location on Earth immune from the effects of gravity." It doesn't matter what the article said, that obviously shouldn't be our title here.)
  - zeta0134 9y ago
    I'd be okay with a very minor change for clarity. Prefixing the title with "Study finds" would make it far less sensational, while staying otherwise faithful to the original article's headline.
    - logicallee 9y ago
      I suggest "No correlation between frequency response and price quartile in 283 headphones". (which fits HN's title limit of 80 characters.) As far as I can see they just judged the quartiles.
- lucb1e 9y ago
  Which is not something we should unconditionally follow. Logicallee makes a good point about that there must be some correlation and the title is kinda ridiculous. I'm not saying we should edit it, I'm not sure whether I'd go that far. But copying titles is a guideline, not a no-exceptions rule.
TekMol 9y ago
I tend to think the title is legit. With the headphones they analyzed, they did not find a correlation.
thaumasiotes 9y ago
Leaving aside the many other problems with your comment, I'd like to draw your attention to this better-informed comment from svantana: > You could, for example, have a bunch of expensive headphones with frequency response that varies randomly in the [-1,+1] dB range, and a bunch of cheap headphones that are in the [-10,+10] dB range -- that would also show up as uncorrelated. > Indeed, they did find a significant difference in magnitude response _error_, although the effect was quite small.
cyphar 9y ago
> Didn't read the link, but could a mod please change this title, which is obviously false? The title is taken from the paper. Whether you agree with the paper's title is not really relevant to the question of what the HN title should be. Because the guidelines expressly state to not editorialise titles. > I skimmed the paper. A better title (for HN) would be "No correlation between frequency response and price quartile in 283 headphones". I don't understand what your point is. Should the paper have tested literally every possible headphone in existence? Also, the paper is actually _not_ comparing price quartiles (though you could be excused for believing that if you only really read the first graph in the paper). The other graphs in the paper are comparing the price of each headphone separately, and show from PCA that the shape of the distribution does not show any correlation between price and deviation from the target frequency response. I actually don't like the way they display distribution data though (just dumping it into a single graph and calling it a day). A much better way of testing their hypothesis would be using GP or similar to try to model the distribution and see whether there was any correlation that way.
- logicallee 9y ago
  Thanks. what was the price of the least and most expensive headphones tested? I tried to figure it out but failed.
  - cyphar 9y ago
    After taking a screenshot and putting into this really cool website that will let you get {x,y} data from plot images[1] (and it's free software). The cheapest headphones they tested were ~$3.95 USD. As an aside, I get really pissed off when researchers don't publish the data they used. It's very frustrating especially when you want to show that you can reproduce earlier results (I've had to use tools like this to include data in plots for my own papers). [1]: http://arohatgi.info/WebPlotDigitizer/ http://arohatgi.info/WebPlotDigitizer/
    - cyphar 9y ago
      Oh, and the most expensive was ~$5k.

navbaker 9y ago

If you read the link, you'll find it's the exact title of the journal article.

logicallee 9y ago
It is still obviously false. It is a sensational title and should be changed so that it is true. (It's like saying, "Scientists find location on Earth immune from the effects of gravity." It doesn't matter what the article said, that obviously shouldn't be our title here.)
- zeta0134 9y ago
  I'd be okay with a very minor change for clarity. Prefixing the title with "Study finds" would make it far less sensational, while staying otherwise faithful to the original article's headline.
  - logicallee 9y ago
    I suggest "No correlation between frequency response and price quartile in 283 headphones". (which fits HN's title limit of 80 characters.) As far as I can see they just judged the quartiles.
lucb1e 9y ago
Which is not something we should unconditionally follow. Logicallee makes a good point about that there must be some correlation and the title is kinda ridiculous. I'm not saying we should edit it, I'm not sure whether I'd go that far. But copying titles is a guideline, not a no-exceptions rule.

TekMol 9y ago

I tend to think the title is legit. With the headphones they analyzed, they did not find a correlation.

thaumasiotes 9y ago

Leaving aside the many other problems with your comment, I'd like to draw your attention to this better-informed comment from svantana: > You could, for example, have a bunch of expensive headphones with frequency response that varies randomly in the [-1,+1] dB range, and a bunch of cheap headphones that are in the [-10,+10] dB range -- that would also show up as uncorrelated. > Indeed, they did find a significant difference in magnitude response _error_, although the effect was quite small.

cyphar 9y ago

> Didn't read the link, but could a mod please change this title, which is obviously false? The title is taken from the paper. Whether you agree with the paper's title is not really relevant to the question of what the HN title should be. Because the guidelines expressly state to not editorialise titles. > I skimmed the paper. A better title (for HN) would be "No correlation between frequency response and price quartile in 283 headphones". I don't understand what your point is. Should the paper have tested literally every possible headphone in existence? Also, the paper is actually _not_ comparing price quartiles (though you could be excused for believing that if you only really read the first graph in the paper). The other graphs in the paper are comparing the price of each headphone separately, and show from PCA that the shape of the distribution does not show any correlation between price and deviation from the target frequency response. I actually don't like the way they display distribution data though (just dumping it into a single graph and calling it a day). A much better way of testing their hypothesis would be using GP or similar to try to model the distribution and see whether there was any correlation that way.

logicallee 9y ago
Thanks. what was the price of the least and most expensive headphones tested? I tried to figure it out but failed.
- cyphar 9y ago
  After taking a screenshot and putting into this really cool website that will let you get {x,y} data from plot images[1] (and it's free software). The cheapest headphones they tested were ~$3.95 USD. As an aside, I get really pissed off when researchers don't publish the data they used. It's very frustrating especially when you want to show that you can reproduce earlier results (I've had to use tools like this to include data in plots for my own papers). [1]: http://arohatgi.info/WebPlotDigitizer/ http://arohatgi.info/WebPlotDigitizer/
  - cyphar 9y ago
    Oh, and the most expensive was ~$5k.