REVIEW
Attempt to distinguish the origins of self-similarity by natural time analysis
Statistical Mechanics · Submitted 25 Feb 2006 · 39 references
Pith's verdict: A genuinely new numerical observation about entropy and power-law exponents, but the missing entropy-estimation details make the central claim uncheckable as written.
T0 · AI review
Why Pith says this
Highlight any text to comment on it.
Referee report
The paper
The paper claims
Shuffling a time series and watching the natural-time variance separates memory-driven from heavy-tail-driven self-similarity, and the differential entropy of that variance peaks at exponent $\gamma\approx1.6\text{--}1.7$.
The reading
What carries the argument
The central object is natural time $\chi_k=k/N$, the normalized count index of the $k$-th event, paired with weights $p_k=Q_k/\sum_n Q_n$ proportional to released energy. The argument is carried by two functionals of the weighted natural-time sequence: the variance $\kappa_1=\langle\chi^2\rangle-\langle\chi\rangle^2$, whose probability density over sliding windows is computed by the procedure of Ref. [17], and the differential entropy $S_I=-\int P(\kappa_1)\ln P(\kappa_1)\,d\kappa_1$ of that density. Shuffling the event order is the operational device that separates the two origins: it destroys memory while preserving the distribution of the $p_k$, so the comparison of $\kappa_1$ in original and shuffled data isolates the heavy-tail contribution.
Extended reading notes
Core claim
In natural time $\chi_k=k/N$, with event weights $p_k=Q_k/\sum_n Q_n$, the variance is $\kappa_1=\langle\chi^2\rangle-\langle\chi\rangle^2$, whose reference value for a uniform distribution is $\kappa_u=0.0833$. The paper's criterion is: if self-similarity comes solely from memory, shuffling the event order makes $\kappa_1$ approach $\kappa_u$ and the entropy $S$ approach $S_u=0.0966$; if it comes solely from the heavy-tailed burst-size distribution, shuffling leaves the most probable $\kappa_1$ essentially unchanged but still different from $\kappa_u$. Applied to synthetic data obeying $P(E)\sim E^{-\gamma}$, the most probable $\kappa_1$ decreases with $\gamma$, and the differential entropy $S_I=-\int P(\kappa_1)\ln P(\kappa_1)\,d\kappa_1$ reaches a maximum for $\gamma$ between 1.6 and 1.7. For earthquakes, $\gamma=1+b/1.5$ converts this to $b\approx 1$, the usual observed Gutenberg-Richter value.
Load-bearing premise
The synthetic power-law data fix the total event count by letting the maximum event magnitude change with $\gamma$, so the entropy curve's peak may be produced by that coupling rather than by the shape of the power-law tail.
Editorial extensions
If this is right
- A self-similar series can be classified by comparing $\kappa_1$ in original and shuffled data: a shift toward $\kappa_u=0.0833$ indicates memory-dominated self-similarity, while a maintained offset from $\kappa_u$ indicates heavy-tail-dominated self-similarity.
- The curve relating $\gamma$ to the most probable $\kappa_1$ gives a natural-time estimator of the power-law exponent from variance statistics alone.
- If the entropy peak is robust, the narrow observed range $1.5\le\gamma\le 2.1$ across many burst phenomena may reflect a common statistical selection mechanism rather than coincidence.
- For earthquakes, the entropy peak implies a Gutenberg-Richter $b$-value near 1, matching the observed global range $0.8\text{--}1.2$.
- Changing the sliding-window length from 6--40 to 6--100 events leaves the entropy curve almost unchanged, so the peak is not an artifact of the chosen window length.
Reading between the lines
- Beyond the paper: the entropy-maximization idea could be tested against other candidate burst-size distributions (log-normal, truncated exponential) to see whether the $\gamma\approx1.6\text{--}1.7$ peak is specific to power-law tails or generic to broad distributions.
- Beyond the paper: because the synthetic protocol varies the maximum magnitude as $\gamma$ changes, repeating the computation with the upper cutoff held fixed would determine whether the entropy peak reflects the power-law tail itself or the cutoff coupling.
- Beyond the paper: the same shuffled-versus-original comparison could be applied to other empirically self-similar records, such as neuronal avalanches or financial bursts, to classify their self-similarity origin.
Original abstract
Self-similarity may originate from two origins, i.e., the process memory and the process' increments ``infinite'' variance. A distinction is attempted by employing the natural time \chi. Concerning the first origin, we analyze recent data on Seismic Electric Signals, which support the view that they exhibit infinitely ranged temporal correlations. Concerning the second, slowly driven systems that emit bursts of various energies E obeying power-law distribution, i.e., P(E) ~ E^-\gamma, are studied. An interrelation between the exponent \gamma and the variance \kappa_1(= <\chi^2> - <\chi>^2) is obtained for the shuffled (randomized) data. In the latter, the most probable value of \kappa_1 is approximately equal to that of the original data. Finally, it is found that the differential entropy associated with the probability P(\kappa_1) maximizes for \gamma around 1.6 to 1.7, which is comparable to the value determined experimentally in diverse phenomena, e.g., solar flares, icequakes, dislocation glide in stressed single crystals of ice e.t.c. It also agrees with the b-value in the Gutenberg-Richter law of earthquakes.
The compiled record
The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
The abstract, sentence by sentence
Self-similarity may originate from two origins, i.e., the process memory and the process' increments ``infinite'' variance. A distinction is attempted by employing the natural time \chi. Concerning the first origin, we analyze recent data on Seismic Electric Signals, which support the view that they exhibit infinitely ranged temporal correlations. Concerning the second, slowly driven systems that emit bursts of various energies E obeying power-law distribution, i.e., P(E) ~ E^-\gamma, are studied. An interrelation between the exponent \gamma and the variance \kappa_1(= <\chi^2> - <\chi>^2) is obtained for the shuffled (randomized) data. In the latter, the most probable value of \kappa_1 is approximately equal to that of the original data. Finally, it is found that the differential entropy associated with the probability P(\kappa_1) maximizes for \gamma around 1.6 to 1.7, which is comparable to the value determined experimentally in diverse phenomena, e.g., solar flares, icequakes, dislocation glide in stressed single crystals of ice e.t.c. It also agrees with the b-value in the Gutenberg-Richter law of earthquakes.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Self-similarity may originate from two origins, i.e., the process memory and the process' increments ``infinite'' variance. | background 1.00 | 0.06 | 0.06 | hedged possibility 1.00 |
| s1 | A distinction is attempted by employing the natural time \chi. | method 0.89 | 0.81 | 0.03 | asserted 0.74 |
| s2 | Concerning the first origin, we analyze recent data on Seismic Electric Signals, which support the view that they exhibit infinitely ranged temporal correlations. | contribution 0.54 | 0.48 | 0.07 | asserted 0.75 |
| s3 | Concerning the second, slowly driven systems that emit bursts of various energies E obeying power-law distribution, i.e., P(E) ~ E^-\gamma, are studied. | contribution 0.54 | 0.54 | 0.03 | asserted 0.98 |
| s4 | An interrelation between the exponent \gamma and the variance \kappa_1(= <\chi^2> - <\chi>^2) is obtained for the shuffled (randomized) data. | result 0.92 | 0.88 | 0.09 | asserted 1.00 |
| s5 | In the latter, the most probable value of \kappa_1 is approximately equal to that of the original data. | result 0.99 | 0.84 | 0.12 | asserted 0.95 |
| s6 | Finally, it is found that the differential entropy associated with the probability P(\kappa_1) maximizes for \gamma around 1.6 to 1.7, which is comparable to the value determined experimentally in diverse phenomena, e.g., solar flares, icequakes, dislocation glide in stressed single crystals of ice e.t.c. | result 0.96 | 0.89 | 0.36 | asserted 1.00 |
| s7 | It also agrees with the b-value in the Gutenberg-Richter law of earthquakes. | result 0.56 | 0.51 | 0.11 | asserted 0.97 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
Figures
Figures reproduced from arXiv:cond-mat/0602603 by the authors.
References
Discussion
Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.