LRydin
commented
2 years ago 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
ghost
I have a question about
In README.md, you argued ``We will fit the first points, since the results are more accurate there'', and used the first points to estimate Hurst exponent. Is there any evidence, such as papers or results of numerical simulations, that you can say so? (I thought this description was intuitively different from the methods described in the original paper of DFA.)