foo_abx - ABXY, what's Y? Options

[bernhold]

I recently learned about ABX tests and wanted to conduct my own, so I installed foo_abx. But there are four buttons, A B X and Y. A and B are the different samples, and X is randomly set to either A or B. But what's Y?

[carpman]

Let's say A = original and B = lossy version:

If X is A, Y is B (and vice versa)

So each attempt you'll be saying A is X or Y (because A is the original).

If you keep getting it right, then you can tell the difference between A and B

C.

[bernhold]

Thanks for clarifying. I now understood what everything is, but I don't understand why I would need Y.

Let's say I listen to A, B and X. After that, I'm sure that A = X. This automatically implies that B = Y. So why would I ever need to listen to Y?

This post has been edited by bernhold: Apr 13 2013, 19:29

[lvqcl]

Let's say I listen to A, B and X. After that, I'm sure that A = X. This automatically implies that B = Y. So why would I ever need to listen to Y?

In this case, you don't need Y.

[carpman]

Because what happens if "After that, I'm" NOT "sure that A = X"?. You might need to listen to Y, then X again, then Y then X again. Sometimes the differences can be very hard to ascertain, and can require a very high degree of concentration.

C.

[carpman]

I should add that you need to do a number of tests for it to be statistically relevant (high confidence that you're not guessing). So, let's say you can hear something in B that isn't in A, you're always looking for that artifact. Well sometimes B will be X and sometimes Y. It switches randomly between the two.

Have a look here for more info:
http://www.hydrogenaudio.org/forums/index....showtopic=16295

C.

This post has been edited by carpman: Apr 13 2013, 19:50

[bernhold]

This is kind of confusing. Let's do an example trial:

A = original, B = lossy
X = lossy, Y = original

I listen to A, B and X in order. After repeating this 10 times, I can't tell if X is A or B. This means I can't distinguish between original and lossy. Knowing that, why would I listen to A, B and Y after that? The outcome will be exactly the same, I won't be able to distinguish them either, as Y doesn't introduce any new info. It's technically redundant.

Could it be that Y is more of a convenience function? Or let's put it this way, if the Y button wouldn't exist, would I still be able to conduct a scientific ABX test?

This post has been edited by bernhold: Apr 13 2013, 20:09

[carpman]

No, it switches.

First trial X = lossy, Y = original
But then next go X may be original etc ...

Look at the link I posted.

Additionally, this is how I would ABX something:

I listen to A then B (many times) until I think I can discern a difference. Then once I've found the thing that makes B different I'll listen to X then Y to see which one is B. Then I'll make a choice, and run the test again about ten times: This will give me a report telling me how I've done.

C.

EDIT: Sorry bernhold, I'm a bit tired, I'm possibly not quite getting exactly what you're asking.

This post has been edited by carpman: Apr 13 2013, 20:30

[bernhold]

No, it switches.

Yeah, I understand that A, B, X and Y are changed after each trial. But my point is still valid, you don't need Y to determine whether X = A or X = B.

I listen to A then B (many times) until I think I can discern a difference. Then once I've found the thing that makes B different I'll listen to X then Y to see which one is B. Then I'll make a choice, and run the test again about ten times: This will give me a report telling me how I've done.

I do it this way: I listen to A and B. In B, I hear a compression artifact. I then listen to X. If I hear that artifact in X, I know that X = B. If I don't hear it, I know that X = A. Why do I need Y? It's confusing, I know, but I still don't get it

Something which just came to my mind: In a single trial, are you allowed to listen to A, B and X as many times as you want? Or are you only allowed to listen to each sample once? If so, maybe this is what I'm missing. Let's say that in one particular trial, you listen to A, then listen to B, then listen to X, then listen to A again, listen to B again, listen to X again. Is that allowed?

This post has been edited by bernhold: Apr 13 2013, 20:56