Fosi Luna 3 Turntable..

[mike70]

Why 75%? And, statistically speaking, what is the threshold (i.e., confidence interval) that defines "more than"? E.g., based on analysis of your data set, is 76% reliably more than 75%, or is 80% reliably more than 75%, or, is it 97%, in a statistically defendable way?

Appearances to the contrary I'm not being persnickety at all. I'm just trying to understand the strength of the conclusion that the cartridges "have an audible difference (statistically a real difference and not a flip of the coin)">

https://www.audiosciencereview.com/forum/index.php?threads/statistics-of-abx-testing.170/

 

[mike70]

https://www.audiosciencereview.com/forum/index.php?threads/statistics-of-abx-testing.170/

as you can see, the 95% of confidence depends in the amount of trials, and i normally do something like 15 (more than 10).

i have challenged people that swear "big differences only a deaf cannot hear" ... and they failed miserably the abx test
A mind changer, a hidden reality.
You can trust your ears, not your brain.

 

[rdenney]

For me, “not subtle” ranging up to “my wife geard it from the kitchen” needs 100% in an ABX test. I would “I can tell them apart maybe 4 out of 5 times” is the very definition of subtle, even marginal.

Rick “especially if one is really concentrating on the test” Denney

 

[mike70]

For me, “not subtle” ranging up to “my wife geard it from the kitchen” needs 100% in an ABX test. I would “I can tell them apart maybe 4 out of 5 times” is the very definition of subtle, even marginal.

Rick “especially if one is really concentrating on the test” Denney

Well, you don't need 100% ... 95% of confidence is statistically right. You need to have a little space to errors, fatigue, etc
We aren't robots
In 10 times, 8 without a difference is equals to no difference... the other 2 times aren't enough to change the result.

 

[rdenney]

Well, you don't need 100% ... 95% of confidence is statistically right. You need to have a little space to errors, fatigue, etc
We aren't robots
In 10 times, 8 without a difference is equals to no difference... the other 2 times aren't enough to change the result.

If being a little test-fatigued is enough to make a difference not noticeable, it’s subtle. We get these night vs. day adjectives from golden-eared audiophiles all the time, and I can tell night from day 100% of the time, even when fatigued.

I didn’t say “no difference.”

Rick “time to make words mean what they say” Denney

 

[mike70]

If being a little test-fatigued is enough to make a difference not noticeable, it’s subtle. We get these night vs. day adjectives from golden-eared audiophiles all the time, and I can tell night from day 100% of the time, even when fatigued.

I didn’t say “no difference.”

Rick “time to make words mean what they say” Denney

I think we're talking about different stuff, I was talking about math... and you're talking about something subjective about tests. It's Ok, opinion isn't debatable.

In hifi (not on a portable mono speaker) I don't find many situations about "night and day" differences really.

 

[antcollinet]

Well, you don't need 100% ... 95% of confidence is statistically right. You need to have a little space to errors, fatigue, etc
We aren't robots
In 10 times, 8 without a difference is equals to no difference... the other 2 times aren't enough to change the result.

Yes they are - a 95% confidence means there is still a 5% chance the "8/10 correct" is just the result of random guessing. It comes down to deciding how important it is for the answer to be correct.


For example - I'd not accept a 5% chance of dying every time I crossed the road.

I'd also agree with @rdenney - if the difference is night and day, you should be able to do 100%. If you are having to strain to hear a difference - or switch back and forth multiple times to detect it - then it is subtle - and of little importance. Certainly not significant enough to impact your listening enjoyment.

 

[morillon]

Fosi Luna 3

Has it been measured a little seriously from the W&F-speed side using known standards?

 

[rdenney]

I think we're talking about different stuff, I was talking about math... and you're talking about something subjective about tests. It's Ok, opinion isn't debatable.

In hifi (not on a portable mono speaker) I don't find many situations about "night and day" differences really.

Me, neither. But we continue to hear about them.

The ABX test is designed to determine if a difference is detectable, and the calculation provides a confidence level of the test statistics. How we apply a confidence level to a decision is a matter of policy, and is not a calculation, nor is it a subjective value. It is a qualitatively determined threshold of acceptance. Some may say a confidence level of 60% indicates a detected difference, but you and I would probably agree that this is still very close to a random result. To claim a reliable ability to detect any difference, people might argue for confidence levels in the range of 75-95% (and my fingers itch with reluctance typing 75).

But I was applying a different threshold for a different test hypothesis, that the difference between them is obvious rather than merely detectable. My reason is that most of the time, audiophiles don’t report barely being able to reliably detect a difference (75-85% confidence). They usually report that every tweak made a transformational difference, and for that claim I’d want 100% reliability. It’s an important distinction for understanding the diminishing returns of a purchase choice.

To the task at hand, I might want to spend hundreds of dollars (or hundreds of dollars more) on a cartridge to get an obvious improvement (i.e. detectable difference—and preference—in every trial). But if the improvement is subtle in an ABX test (i.e. undetectable in some trials for whatever reason), it may make no useful difference commensurate with cost in practice, and my money would be better spent elsewhere.

Rick “works with statistics every day” Denney

 

[RoA]

Absolutely, as the 95ML/SH should wipe the floor with the soggy old Rega pickup if sound quality closer to truth is concerned (the Bias/Elys/Exact pickups were bad when they first appeared in 1987 or so, but the dealers who knew no better sold 'em regardless).

I changed the VM conical that came with my 100C to the VM95ML. The difference is small. That either makes the VM95C an incredible bargain or the the VM95ML hyped (imho)

I will probably try an AT VM745xML.

 

[DSJR]

I changed the VM conical that came with my 100C to the VM95ML. The difference is small. That either makes the VM95C an incredible bargain or the the VM95ML hyped (imho)

I will probably try an AT VM745xML.

You should hear the difference on cymbal-work especially, as the conical tip kind-of closes up subjectively. A stock AT91/VM95C is great from mid bass to mid-midrange, but firstly, going elliptical *should* open the 'window' wider further up the range. The 95ML sharpens it up rather more up-top I found (tracking at 2g) which can work so well with some LPs especially, the SH seeming to relax the 'tension' just a little but without losing the extra details. All subjective and maybe personal to me, but I share anyway, as vinyl is rather more than a technical appraisal (why many here simply don't go there )

The 500/700 body does have refinements internally over the humble 95 series and this at least, should improve separation figures and so on. The posher internal wiring and terminal pins may or may not make any difference at all, but it makes one feel better knowing it's there I think. The metallic mount also seemed to relax the hf response just a little, this in an otherwise 'inert' Rega tonearm (I think the old Planar 9 was used as test bed at the time, but Lowbeats.de, which did these tests and sound-samples some years back, don't seem to do tech tests on cartridges any more sadly)

 

[mike70]

Me, neither. But we continue to hear about them.

The ABX test is designed to determine if a difference is detectable, and the calculation provides a confidence level of the test statistics. How we apply a confidence level to a decision is a matter of policy, and is not a calculation, nor is it a subjective value. It is a qualitatively determined threshold of acceptance. Some may say a confidence level of 60% indicates a detected difference, but you and I would probably agree that this is still very close to a random result. To claim a reliable ability to detect any difference, people might argue for confidence levels in the range of 75-95% (and my fingers itch with reluctance typing 75).

But I was applying a different threshold for a different test hypothesis, that the difference between them is obvious rather than merely detectable. My reason is that most of the time, audiophiles don’t report barely being able to reliably detect a difference (75-85% confidence). They usually report that every tweak made a transformational difference, and for that claim I’d want 100% reliability. It’s an important distinction for understanding the diminishing returns of a purchase choice.

To the task at hand, I might want to spend hundreds of dollars (or hundreds of dollars more) on a cartridge to get an obvious improvement (i.e. detectable difference—and preference—in every trial). But if the improvement is subtle in an ABX test (i.e. undetectable in some trials for whatever reason), it may make no useful difference commensurate with cost in practice, and my money would be better spent elsewhere.

Rick “works with statistics every day” Denney

I get your point now

I started from the point of "there's no night and day in hifi", much more blah blah than real differences.
My 90-95% confidence it's more than enough to convince me (and yes, that's subjective )

 

[mike70]

You should hear the difference on cymbal-work especially, as the conical tip kind-of closes up subjectively. A stock AT91/VM95C is great from mid bass to mid-midrange, but firstly, going elliptical *should* open the 'window' wider further up the range. The 95ML sharpens it up rather more up-top I found (tracking at 2g) which can work so well with some LPs especially, the SH seeming to relax the 'tension' just a little but without losing the extra details. All subjective and maybe personal to me, but I share anyway, as vinyl is rather more than a technical appraisal (why many here simply don't go there )

The 500/700 body does have refinements internally over the humble 95 series and this at least, should improve separation figures and so on. The posher internal wiring and terminal pins may or may not make any difference at all, but it makes one feel better knowing it's there I think. The metallic mount also seemed to relax the hf response just a little, this in an otherwise 'inert' Rega tonearm (I think the old Planar 9 was used as test bed at the time, but Lowbeats.de, which did these tests and sound-samples some years back, don't seem to do tech tests on cartridges any more sadly)

I didn't find great differences between vm95ml and vm540. Between vm95e and vm95ml... yes, notably in tracking capabilities and high frequencies

 

[egellings]

The following gut punch part is when they're telling you the price.

Yes, drill sergeant! Right away, drill sergeant!