From: naimead on 5 May 2010 08:31 These data isn't part of an unknown distribution.I have an original signal with a non-Gaussian distribution and I passed it through IAAFT() to obtain 39 surrogates(which theoretically share the same unknown non-Gaussian distribution).Then I calculated its one's slope asymmetry and these are the 39 values I displayed in my previous post.I just want to prove if the original and surrogate data are statistically significant but I can't do this with the classical 1.96 value since they don't follow a normal distribution. So I need to see if the slope asymmetry of the original data is larger than the 95-percentile of the unknown distribution of the slope asymmetries of the 39 surrogates. If this is the case then my original and surrogate data are statistically significant. This technique is applied in the following paper where the author using mle found that his unknown distribution follows a gamma distribution:"Nonlinearity testing in the case of non Gaussian surrogates, applied to improving analysis of synchronicity in uterine contraction"
From: dpb on 5 May 2010 13:04
naimead wrote: .... > Is it possible to calculate 95-percentile of a non-Gaussian > distribution without having to define its exact distribution? .... There's always Chebysev's inequality (altho quite conservative for many distributions)... -- |