Hello, funny what the thread became. To me it is much more pleasant than that ruddy old can. opaqueice;229901 Wrote: > Hi Jay, thanks for the explanation. I teach physics and am also pretty > familiar with probability and statistics. I was asking about sequence > to make sure you hadn't thrown away some initial series of > not-so-successful trials. People often do that, thinking they need to > "warm up", but it can be extremely misleading because you may (perhaps > not consciously) throw away results until you achieve something > consistent with your expectations. That's sometimes called > cherrypicking, and it's a big concern in science (it's one of the > reasons you should never fully trust the reported statistical > significance of experiments). > It is the reason why we should expect dozens of studies disprooving the glorious research hypothesis laing arround in draws never being published. Why is science such a lengthy and costly untertaking. Because they go on for years until they accidentally modified their original experiment until they finally hit a difference that the can shout arround about. Nope. I took ten songs and after that was not only statistically but subjectively quite sure that there actually is a difference in sound audible. I would say transporters rca makes for a more solid bass and mids i.e instruments that seem to jump at me a little closer. > > Do I understand you correctly that you identified the source correctly > 8 times out of 10 trials? Yes.
> > The chance of scoring 8 or higher out of 10 randomly is 5.5%. However > in this instance one should double this, because the null hypothesis > (that the two sources are identical) would be rejected either by a very > high score or a very low one (in other words you should use the "two > tailed" distribution). > Not sure about that one. I calculated the distribution which aproximates normal gaussian curve (sorry my english is not good enough) VERY roughly out of my head as follows. For the ten categories I ended up with Right trials: 0 1 2 3 4 5 6 7 8 9 10 1 10 45 120 210 252 210 120 45 10 1 = 1024 total events That means that there are 56 paths where I get 8 out of 10 right? Now why should I if I confuse the two sources completely i.e less than 2 right trials assume that they are different? Because that would indeed make two tailed distribution applicable. > > 8/10 or better OR 2/10 or worse will occur randomly 11% of the time, > which doesn't meet the criterion of stat significance. That's > especially important here, where the test was single blind (your > girlfriend and you were in the same room, right?). And why is the test single blind then? I heard two songs from sources randomly choosen i.e first second second first first first second second and then gave my judgement without seeing either the display of the slim device or my Marantz amp. -- jaysung ------------------------------------------------------------------------ jaysung's Profile: http://forums.slimdevices.com/member.php?userid=12375 View this thread: http://forums.slimdevices.com/showthread.php?t=38653 _______________________________________________ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
