Listening Tests & The Bell Curve

[Primalsea]

Split from: Who Will Win - http://theartofsound.net/forum/showthread.php?t=4715&page=6


Every now and again I have a fit of something or another and begin to talk sense! It doesn't happen often.

I agree somewhat with Dimitri but we are starting slide a little bit into another argument; Determinism vs Randomism.

Why should it be that everything can be explained? It shouldn't stop us looking to see if it can but we have to accept that maybe at the end it can't.

Here's something to think about, I have never seen results of any listening tests anywhere, ever, double blind or otherwise represented as a bell curve.

Why is that??

By the way this is not really related but it just fell into my head from the dimension where thoughts are born.

[Marco]

If I knew what a 'bell curve' was (and what it represented) I might be able to offer an opinion...

I know what a bell-end is, if that helps? :eyebrows:

Marco.

[REM]

If I knew what a 'bell curve' was (and what it represented) I might be able to offer an opinion...

I know what a bell-end is, if that helps? :eyebrows:

Marco.

Bell curve a.k.a. normal distribution curve = statistical representation of any characteristic within a given population or sample. The larger the sample the more symmetrical the curve, which plotted onto a graph resembles a bell, see HERE (http://www.netmba.com/statistics/distribution/normal/).

A bell end typically post on forums without re-checking the text formatting....:rolleyes:

[Themis]

Every now and again I have a fit of something or another and begin to talk sense! It doesn't happen often.

I agree somewhat with Dimitri but we are starting slide a little bit into another argument; Determinism vs Randomism.

Why should it be that everything can be explained? It shouldn't stop us looking to see if it can but we have to accept that maybe at the end it can't.

Here's something to think about, I have never seen results of any listening tests anywhere, ever, double blind or otherwise represented as a bell curve.

Why is that??

By the way this is not really related but it just fell into my head from the dimension where thoughts are born.
I have a (rather) accurate view of the statistical analysis and why/when/how it can be used for listening tests. I need some time to explain, though, I'll have more time tonight.

What I can rapidly say, nevertheless, is that ABX is statistically flawed (I have a solid argumentation proving it) and that it's more accurate to use conventional testing methods, rather than statistical analysis for most audio tests. You just need to be careful with some common "testing traps". ;)

More later.

[Primalsea]

Now hifi testing results represented as a Bell End, that's something worth seeing.:)

If I understand correctly if you had a proper methodology and asked people to respond to some hifi set ups the result should show a bell curve when plotted on a graph. apparently its a sign that the test is well conceived. However I don't know enough about it other than to ask why you don't see it in hifi related tests.

[Themis]

Now hifi testing results represented as a Bell End, that's something worth seeing.:)

If I understand correctly if you had a proper methodology and asked people to respond to some hifi set ups the result should show a bell curve when plotted on a graph. apparently its a sign that the test is well conceived. However I don't know enough about it other than to ask why you don't see it in hifi related tests.
A bell graph is a sign that an event is evenly happening within a standard deviation.
You cannot see it in hifi related tests, simply because you need to track only one event : usually hifi tests don't involve repetition of the same event (with one or several auditors).

[Marco]

Hi Dimitri,


What I can rapidly say, nevertheless, is that ABX is statistically flawed (I have a solid argumentation proving it) and that it's more accurate to use conventional testing methods, rather than statistical analysis for most audio tests. You just need to be careful with some common "testing traps".

More later.


Yes please!

Can we have this on a separate thread? :)

Marco.

P.S I'm still waiting to see pics of the ART digital lead you promised in your gallery. Just to remind you like... ;)

[Primalsea]

Is that because its just impossible or is it just because the tests normally carried out on hifi dont allow it? Could a valid test for hifi be created?

Could it be possible that a groups of people rank and number of systems on several points, relative to each other system, say 1 to 10 or something, with one of the points being preference?

What I'm trying to get at is ultimately I would expect there to be distinct groups of people based on how they would rank particular aspects. Further these groups would have a preference are particular equipment.

[Themis]

Is that because its just impossible or is it just because the tests normally carried out on hifi dont allow it? Could a valid test for hifi be created?

Could it be possible that a groups of people rank and number of systems on several points, relative to each other system, say 1 to 10 or something, with one of the points being preference?

What I'm trying to get at is ultimately I would expect there to be distinct groups of people based on how they would rank particular aspects. Further these groups would have a preference are particular equipment.
Yes, it could be possible. Simply, you need to track a unique event each time.
It could be a unique preference point, for instance, indeed.

I guess it doesn't happen, perhaps because it's difficult to find someone to organize it :
- The reviews prefer a single experienced reviewer (or two)
- The manufacturers don't have an interest into comparing their equipment with a single competitor.

[Alex_UK]

I know what a bell-end is, if that helps? :eyebrows:


You could even live there, if you wanted!

[Marco]

I think that I might make that into my avatar..... :eyebrows:

Marco.

[Alex_UK]

Ended up there purely by accident with a car club in the summer... we did laugh!

[The Grand Wazoo]

Just to clarify for those who still don't know what a bell curve is
Take a look at this:

http://foper.unu.edu/course//wp-content/uploads/2007/02/fige10.jpg

This is a timber yield curve - we use them in forestry a lot & this one shows a way to determine the most economically advantageous time to harvest a crop of timber.

The 'bell' is the green line (it looks like half a cross section of a bell). You get this in a statistically normal population when measuring all sorts of parameters. A simple example would be where you're recording ages - in any given population you'll have a wide range of ages & the bulk of the community will be in the middle of the curve, while it'll tail off to the lower end of the scale and also to the higher end of the scale.

To relate that to the curve above, the optimum economic advantage can be gained where you cut down trees at the tipping point before the annual growth begins to slow (the point of maximum mean annual increment). If the trees had been growing quicker the centre area of the curve would be steeper & vice versa.

There you go my friends, today you've learnt something about:

hi-fi,
statistics,
the yield function of crops and
how to cut down your very own rain forest!!