Hurst exponent approx.

[pmp.pp...@gmail.com]

I am attempting to approximate the hurst exponent using pwt and I am having some issues, I'm not sure if my calculation is incorrect or there are perhaps some numerical issues(doubt this).

The code is below:

y = np.random.randn(20)

# This is another approximation to be sure the wavelet approx is close

h0 = eg.hurst(y)

w = pywt.Wavelet('db4')

c = pywt.wavedec(y, w)

levels = len(c)

D = []

for i in range(len(c)):

   dj = 0

   l = len(c[i])

   for j in range(l):

       dj += float(pow(c[i][j], 2))

   D.append( dj / float(l) )

print D

slope,intercept=np.polyfit(np.log2(D),range(len(D)),1)

hurst = abs((slope-1)/float(2))

print(h)

print(hurst)

[Robin Xu]

http://www.tandfonline.com/doi/full/10.1080/17446540701367444#preview   This website may help you 

在 2013年2月6日星期三UTC+8下午12时23分45秒，pmp.pp...@gmail.com写道：

[ksandr...@gmail.com]

Sorry, did you find the answer? 
I trying to use pywawlets to find the Hurst exponent, but have different numbers with different algorytms.

среда, 6 февраля 2013 г., 7:23:45 UTC+3 пользователь pmp.pp...@gmail.com написал: