I've just finished an initial release of a new ABX comparator. It's a preliminary alpha version, so don't be very hard with the look of the program or some lacking features.

It's at www.kikeg.arrakis.es/winabx/winabx.zip

Any comments or suggestions are welcome.


Edit: updated download path.

Also, I'm aware that there are same bugs related to not properly opening a wave file at the beginning of the program, I'll fix them when I have some time. Also, don't get mad trying to change A and B wave files once the program is running, with this version now you just have to close the program and run it again to select new wave files. Sorry

Thanks.

Cool. One comment:

It seems like there are too many buttons. Is there really a need to have both X and Y present? Especially since it seems like X is always different from Y.

ff123

Yeah, Y is the opposite of X, I just put it there to make the tests more (in my opinion) flexible. You know, you can compare X with Y and decide which is which, just like an blind A/B test. I think it helps (at least to me) with subtle differences.

Also, I'm using your code to compute the p values, when I make some kind of documentation I will give you credit

hi. all

who knows if there is a softwre which can testing two codec with CMOS method?

how to make ANOVA analysis?

thanks.

What is the CMOS method?

ff123

I've put into the .zip file an Excel table with binomial distribution calculated p-values up to 100 trials.

I don't want to be rude, since i think you prog is a good initiative, but it has nothing in common with abx anymor.
The purpose of abx, is to compare a mysterious x with a known a and a know b.
Yours could be called an "ABXY", where you can cheat by comparing X and Y (yes it's not really cheating, but i think it introduce a disturbing which has nothing to do with the original idea of abx, and the scientific methodology it relies on.

Yeah, it's more an ABXY test.

But I don't see why it can be so disturbing. The point is to be able to identify a file in a blind fashion. There's nothing at the metodology used that can spoil this. If you don't like the "Y" option, just don't use it. I use it because I find it useful, and think it can be useful for others too.

With the "Y" option, I've just added the option to optionally perform the equivalent to a blind A/B test, or an ABC/HR test, or a mix of all them.

It seems scentific to me as well.
I don't see how someone could recognize X/Y vs A/B without failing if he can't objectively make the difference in the first place.