This is an implementation of the Hurst exponent calculation that is smaller, simpler, and quicker than most others. It does a dispersional analysis on the data and then uses Matlab's polyfit to estimate the Hurst exponent. It comes with a test driver that you can delete.
I think the code doesn't run correctly on R2016a. I find the code of genhurst in http://www.mathworks.com/matlabcentral/fileexchange/30076-generalized-hurst-exponent. gives more precise result, but the wfbmesti command in wavelet toolbox is precise and easy to use .
I got some H greater than 1 when i ran the code. Could you tell why? thx.
This code works fine, but it is meant for FGN,if correct result is required for FBM, then take the differentiation and check the result.
FBM = wfbm(0.8,1000);
rand, generate gaussian random noise rather than Brownian noise
Tomaso Aste has uploaded (18 Jan 2011) a code there : http://www.mathworks.com/matlabcentral/fileexchange/30076-generalized-hurst-exponent.
following the test proposed by hau and sen_saven gdsgds :
genhurst(test) --> 0.7934 on a 7.9 R2009b 64bit
genhurst(test) --> 0.7926 7.0 R14 32bit
the code refer to T. Di Matteo et al. Physica A 324 (2003) 183-188 and has no rating until today (04/04/11)
why do I get an error when trying to use "function" in the way this author does? I am running matlab 2010
I am looking for matlab code for calculation of Hurst index as well..Please let me knw if you come across any reliable code..The code at this link looks interesting
well...same problem as hau
testing Hurst calculation
Hurst exponent = 0.98
and even when taking the diff I get
Hurst exponent = 0.71
not 0.8 as it should be..
Is there a matlab code for the calculation of H that gives precise results?
hau, you need to apply the code to the diff of your data. Then it gives the right answer.
I think the code doesn't run correctly on R2008b. No matter which kind of data I plug, I got ~0.96. For example, I used wfbm command in wavelet toolbox to generate a fBm with H=0.2 with length 100000, the result by this code is 0.96; while H=0.8, the result is the same. However, the command wfbmesti in wavelet toolbox can give me very correct answer. Does anyone know what's wrong?
It doesn't really matter what base we use, afterall the slope will be the same.
correct? base for log is '10', not '2' ?
this is good software