- Joined
- May 21, 2019
- Messages
- 4,036
- Likes
- 6,833
Haven't seen anything like that on their website or their posts here.For those upset with the marketing of Uptone, I’ll bet you my bank account that the folks at Matrix would tell you their top range DAC sounds better than their budget range despite both measuring below the audibility range for distortion, etc.
I took a look. There are some issues with the measurements that I won't get into as their conclusion was correct.
On blind test, unfortunately there is no conclusion you can draw from any of it. Imagine if you said you could predict which way a coin toss will go. You predict head, I flip the coin and get head. Do I conclude that you can predict the coin flip? Of course not. You had 50% chance of being right by guessing. The standard in statistical analysis is to rule out 95% probability of chance. That would require 10 flips and you predicting 8 of them correctly. You could still be guessing but at least we have high confidence that likely are not.
Here, the test is more complicated. A number of switches were tested with a number of different test tracks. Somehow we need to verify if preference for certain switch rises up above level of chance per above. We could also do the same with respect to what music was played (these are the "variables" in statistics parlance). This type of analysis is called Anova or analysis of variances invented by Fisher. Every blind test published formally has this type of analysis. Without it, the results cannot be trusted any more than we can trust the first two coin flip guesses from you.
We also need to find testers who were clearly randomly guessing and as such, diluting the value of the rest of the tests. It may turn out that most of them in which case, there would be nothing left to rely on! This is called a "t-test"
I can perform such an analysis but we need the data. They seem to be in the graphs but I don't have the time to sit there and type them in. If someone wants to do that, we can perform the analysis. Here is what the analysis looks like (from a paper on a lossy codec):
View attachment 41511
We strive to keep P-level less than 0.05.
Not really , your ' point ' has been addressed but you have chosen to just repeat it over and over rather than expand on your argument.
You had the nugget of a good position, maybe I will play devils advocate next time ha ha
You get a c-
Tip , have a read of the forum there's some honest introspection from certain quarters that would enhance your argument if you were to use it.
What's the difference between what Matrix Audio and Uptone Audio are really doing?
Naahhhh .... that's not going to happen. Their "beef" is with UpTone
Specific point #1 - If there are any sig. Interactions the Main Effects cannot be interpreted in isolation.
Specific point #2 - Any sig. interactions involving >3 vars. would be next to impossible to interpret (by no matter whom or how clever).
Specific point #3 - This data set will lack statistical power (severely) and, so, is academic anyway wrt post hoc tests (if that's what you meant by t tests although t tests usually means something else).
Amir's tests are such a "blessing" and "revelation", "separating the wheat from the chaff", compared to all the subjective praises for expensive devices (which praises are sometimes even bought by money or by offering expensive devices, e.g. as is the case with some "professional" reviewers).
He's in the forum business, customer meet retailer , retailer meet customer. Everyone feel seen and apart of something bigger whilst being entertained.I just read the threads on this thing over a Audiophool Style, and I have to say the head of that place is totally biased for Uptone.
They are probably his best advertiser. Still, the grovelling is rather over the top on this one.I just read the threads on this thing over a Audiophool Style, and I have to say the head of that place is totally biased for Uptone.
What? I read everything up to this point and figured I give you the assignment and then you tell us this? There is not a lot of data here so we could do some analysis in Excel.Specific point #4 - I don't have SPSS and I'm not volunteering to crunch the data that are there.
You’re making my point for me. That razor thin line of audibility is pretty shaky ground to exalt one company and send the other to the proverbial guillotine.
You made me look at that now closed thread and saw this from Chris:I don't want, intend, or mean to get bent out of shape, and I like CA/AS and I like Chris but goddamn he's attempting to make points and he can't even be bothered to do the cursory research about what I and others are speaking about:
Just to give an idea of the robustness required for such tests:
Without checking exhaustively, if you *had* to map Analysis of Variance to this:
One DV = SQ [or favourability] score - "continuous" for purpose of [M]ANOVA
IV #1 (Between Ss) = Switch (with s levels where s = # switches) [We are looking for a (statistically sig.) Main Effect here - Or not - /joke Science is like the spiritual life - You tend to find what you're looking for - sooner or later]
IV #2 (Between Ss) = Music (with m levels where m = # different Music tracks played)
IV #3 (Within Ss) = Rater (with r levels where r = # Raters) [presumably every S rated every Switch and every Music track]
So this is a s x m x r [at least] "mixed" or "split plot" Design since there is a blend of Between and Within Ss vars..
Specific point #1 - If there are any sig. Interactions the (constituent) Main Effects cannot be interpreted in isolation.
Specific point #2 - Any sig. Interactions involving >3 vars. would be next to impossible to interpret (by no matter whom or how clever).
Specific point #3 - This data set will lack statistical power (severely) and, so, is academic anyway wrt post hoc tests (if that's what you meant by t tests although t tests usually means something else).
Specific point #4 - I don't have SPSS and I'm not volunteering to crunch the data that are there.
Specific point #5 - Like lawyers statisticians may disagree (about how to legitimately process a given data set).
Specific point #6 - There any many methodological considerations that could be scrutinised.
Specific point #7 - This test didn't include the ER.
General point - The realm of objective vs. subjective (and relationship of any hypothetical test to ER experience) is fraught with interpretation hazards of all kinds and we need each other in the collegiate sense if we wish to gain real understanding - too much emphasis on individuals no matter whom is unhelpful.
I could do more but this is on the back of a fag packet. I'm not arguing - I'm going to bed. I'll have a look tomorrow, and if I've fouled up significantly (at p<0.05) I'll admit it.
There was NOT with Matrix i as that has streaming built-in. So I used the EtherRegen direct to it. And then used my PC streaming content to it (and hence all the bits went through ER). So to be abundantly clear, ER fed the Matrix i DAC directly.Was there a computer between ER and DAC on the test you did, Amir?