Which Component Cables (Need answers fast) (1 Viewer)

[Mike Knapp]

quote: If someone does seem to do this, tricking them by unplugging and plugging in the same cable set will surely trip them up. [/quote]
Why would you want to resort to trickery to "trip someone up" because they were getting 100% correct results?
It seems as though you arent interested in finding the truth at all but rather just dis-proving something at any cost.
Mike
[Edited last by Mike Knapp on October 26, 2001 at 09:00 PM]

 

[Jim A. Banville]

"Why would you want to resort to trickery to "trip someone up" because they were getting 100% correct results?
It seems as though you arent interested in finding the truth at all but rather just dis-proving something at any cost.
Mike"
Not at all
If someone truly sees a difference when switching back and forth between cable "A" and "B", I want to TEST whether he really sees a difference or if he just thinks he does. If the subject says he sees a difference between "A" and "B" 100% of the time, shouldn't he also see "no difference" when he thinks you are switching cables, but actually aren't? If you have someone seeing differences when he is watching the same cable, HE GOT PROBLEMS! If he properly identifies that he is watching the same cable after he was previously 100% right that he was watching two different cables when he really was, then I have MUCH more confidence that he is REALLY seeing a diference. Simple as that. Just testing the testers

 

[Guy Kuo]

AB'ing the cables implies a switch from one cable to the other at each sampling point. If the cadence is not broken up then even with the same cable it possible for an observer to lock into a sequence of better, worse, better, worse. You have to randomize which cable is being viewed so patterns caused by the cadence are removed and you only have data regarding differences seen by the observer. The other requirement is separation of the person administering the test from knowledge of which cable is being switched in for the observers to check.
One way to perform this is to have two cables (alpha and beta). These are connected to a rapid switching system so observers can see an nearly instantaneous switch from A to B. At each sample point, which cable is A and B gets randomized. The person performing the switches of the cables must be not seen or heard by the observers. This is to eliminate any unconscious clues being given to the observers as to which cable is which. Facial expressions, body language, voice, stance, and even the sound of the switch or connectors being moved in one direction vs the other could all yield subtle clues to the observers. The clues must be eliminated or else you contaminate the data.
Because the way trial results are recorded can also bias if the recording person knows which cable is which, the information recording person must not know the cable sequence nor have any contact with the person switching the cables. Otherwise, you could have the equivalent of the recorder smiling every time the observers begin to choose a specific cable.
Only at the conclusion of multiple comparison trials between A & B is the randomization of cable assignments revealed to the observers and recording person. By following this isolation of of information and unconscious clues, one can measure in an unbiased manner whether or not the observers actually prefer one cable over another based upon the cable performance alone. At that point the study has been unblinded and the data crossed against the randomization order to calculate if there is a difference of statistical significance. Speaking of significance, the smaller the difference the more trials are going to be needed to yield a statistically meaningful result.
The point of all this is that a truely clean and fair double blind study requires more details than most of us realize. Those "double blind" tests that get put together in the industry or by magazines may well not stand the scrutiny of someone like an epidemiologist who actually specializes in understanding and designing studies.
------------------
Guy Kuo
Link Removed
Ovation Software, the Home of AVIA DVD
[Edited last by Guy Kuo on October 26, 2001 at 10:54 PM]

 

[Mike Knapp]

Jim,
I assumed you were using a random method in the first place, like a coin flip or something. Even a "test fighter" like me wouldnt run the test in a predicitable manner such as A-B-A-B-A-B-A-B.
I thought you were suggesting running a proper (random) test and then when someone got it right, trying to trick them in some way. The test (if it is done correctly) must be random, which would indeed include several instances where the wires were not changed at all.
Mike

 

[Jim A. Banville]

I would start off with A-B-A-B, but then I'd switch to B-B-B-B, then A-A-A-A, then back to A-B-A-B, or something similar. *IF* there is a visible difference between the cables, a "videophile" should see it 100% of the time. I would have two tests run in silence, with the observers wrting down whether they see a difference between cable A and B in the first test, then which cable they think is "better" in the second test.
Regarding not seeing the switcher, my idea about placing the DVD player on the floor (or a short stand) behind the TV would solve that. I wouldn't add a switching unit into the test because that would just give the "loser" an argument as to way the test didn't work out as they thought it should.
I'm not talking about a test so perfect that it would stand up against the most stringent scientific analysis. I'm talking about a simple un-baised test that would tend to solve the question- "are there visible differences between cheap and expensive cables?" If the test turned out showing that observers were 100% correct in their ability to see these differences, and in fact could pick out a "better" cable, and that cable was a premium cable, I would shut my mouth and NEVER sway anyone away from expensive cables In fact, if the cable was in my price range I'd consider buying it!

 

[Guy Kuo]

A test so "perfect" IS required to remove the tester and observer biases. Anything less leaves the contamination in the results. If you serve as the cable switcher, randomizer, and result recorder you can and will contaminate the observer reactions with your own biases. That's fine if the results are really gross as in yep he's dead vs nope he's still talking. When the differences are more subtle, the noise created by the testing methodology threatens to swamp out the desired data. Cable comparisons are fraught with agendas on the part of nearly everyone who gets involved. One has to either remove the human bias factor by going entirely instrumented for all data or reduce it by taking away the ability of preconceived notions to alter the results.
------------------
Guy Kuo
Link Removed
Ovation Software, the Home of AVIA DVD

 

[Jim A. Banville]

As long as the observers use "ballots" that have something like "second view- same/different", where they circle their choice, there could be no doubt their intent. I do not recommend counting dimples or hanging chads And seasoned citizens from W. Palm Beach County are banned from the test!