question: About estimation of hurst exponent. about mfdfa HOT 5 CLOSED

Dear @yukki-jpn, you caught a very good point – which is in fact incorrect! This is indeed described in the original paper¹, the authors explicitly state that the optimal lag to fit the fluctuation function is somewhere between lags > 10 points and lags < N/4, with N the total number of datapoints.

I am working already on changing this in the README and I have to surely do the same for the documentation, I don't want to lead the reader to something that is incorrect. The results one obtains using short segments are not necessarily wrong, but there are cases where they do lead to wrong estimation. Overall, MFDFA tends to underestimate the Hurst coefficients in purely fractal timeseries.

¹Kantelhardt, J. W., Zschiegner, S. A., Koscielny-Bunde, E., Havlin, S., Bunde, A., & Stanley, H. E. (2002). Multifractal detrended fluctuation analysis of nonstationary time series. Physica A: Statistical Mechanics and Its Applications, 316(1-4), 87–114

from mfdfa.

Dear @LRydin
Thank you for your rapid reply.
``I am working already on changing this in the README and I have to surely do the same for the documentation,''
I am looking forward to that modification.

We are currently writing a paper on the long memory process in the financial market, and we intend to use your program and quote your paper. We, therefore, need some peer reviews of our statistical procedure on the estimation of the Hurst exponent. Could you help us with peer review of our paper?

from mfdfa.

Dear @yukki-jpn, I'm happy to help in anyway I can :).

from mfdfa.

Dear @LRydin

We are glad to hear such a reply. We want to get in touch with you using email. Could you let me know your email address if you disclose your email address? Otherwise, please send me an email at s2120443(at)s.tsukuba.ac.jp (please replace (at) into @). I think either I or Dr. Kanazawa will make contact with you in a few days.

from mfdfa.

You can find me at [email protected] :).

from mfdfa.