darrenyeats wrote: > The problem with this statistical angle is that a large number of > samples is needed.
Hi Darren! I'll have to take issue with you here. The whole point of doing statistical analysis on data samples is that the samples do not need to be particularly large to yield statistically significant results, and in fact increasing sample size is of diminishing value compared to the effort required to achieve it. Many sampling schemes use fixed sample sizes to make predictions about underlying populations of considerably variable size. I haven't fully researched the statistical angle yet, but given that the identification of "X" as "A" or "B" has a 50% individual chance of being correct by chance I was thinking along the lines of a "Normal" approximation to a binomial distribution to see how many standard deviations (right or wrong) each listener's answers are from the expected value which would be that you could expect to get half of your answers right using ear defenders & a fair coin toss with "A" for heads & "B" for tails, or vice versa. Perhaps someone would help me out if I'm up a gum-tree with this approach. In any event, once the results have been collected, statistical analyses of any kind can be applied later. Dave :) ------------------------------------------------------------------------ Golden Earring's Profile: http://forums.slimdevices.com/member.php?userid=66646 View this thread: http://forums.slimdevices.com/showthread.php?t=106914 _______________________________________________ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/audiophiles