{"id":"57c81490-137f-40be-af9c-ad1e9ac082eb","arxiv_id":"2608.08666","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"For three very different time series, q-Gaussian fits from histograms vary strongly with bin number, while a bin-free CDF fit returns a single stable q value for each system.","lead":"This paper compares three ways to estimate the heavy-tail parameter q in non-Gaussian systems, using solar wind, Bitcoin, and river flow data. It finds that histogram-based methods give unstable results that depend on bin count, while a cumulative distribution fit avoids binning and gives more consistent estimates.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"CDF method's reported q values are unvalidated: Eq. (9) is not the CDF and its prefactor/normalization are inconsistent, so the central bin-free stability claim lacks support.","rationale":"The paper's headline result has two parts: histogram-based methods are bin-sensitive, and the CDF method is bin-free and more stable. The first part is reasonably supported by Figures 7-9, which show bin-to-bin variation. The second part requires that the CDF fit is a correct estimator of q. This is the load-bearing condition because every subsequent conclusion about the three systems uses the CDF q values. That condition is not secured: Eq. (9) is structurally a normalization integral rather than a CDF, the constants in the printed formula appear inconsistent with Eq. (1), and no code or synthetic-data validation is provided. A single number per dataset is trivially stable; stability in the sense of accuracy or estimator quality requires ground-truth checks. The reader's weakest assumption targeted the CDF implementation, and I agree that is the weak point, but the specific sign-divergence argument does not survive inspection. The more defensible issue is the incomplete CDF expression and prefactor/normalization errors. Because the correct fix is well-defined and the central claim would be supported if the CDF pipeline passes synthetic validation, the reader's conditional verdict remains appropriate; no stronger rejection is warranted on the evidence in the manuscript.","tokens_in":7129,"tokens_out":8769,"duration_ms":87538,"concrete_test":"Generate synthetic samples (e.g., n=10^5) from q-Gaussians with known q in {1.5, 1.7, 2.0, 2.5} and fixed b, then apply the manuscript's CDF fitting routine. Independently compute the theoretical CDF from the closed form F(x)=1/2+(1/2) I_{b x^2/(q-1)}(1/2, (3-q)/(2(q-1))) and refit q. If either recovered q deviates from the true value by more than 0.02, the reported CDF estimates are not validated and the binning-stability conclusion is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim, that the CDF fit yields a single, stable q, requires the theoretical CDF used in the fit to be correct. This is not established. In Section 3.3, Eq. (9) integrates from 0 to infinity, so it is the normalization integral, not the CDF F(x)=∫_{-∞}^{x} p(x')dx' defined in Eq. (8); no incomplete-beta CDF expression is given. Moreover, substituting μ=x^2 and ν=1/(q-1) into Eq. (1) gives p(x)=a(1+(b/ν)x^2)^{-ν}, so the half-line integral is (a/2)(ν/b)^{1/2} B(1/2, ν-1/2). Eq. (9) instead places a^ν in the denominator, and its second equality drops the (ν/b)^{1/2} prefactor, writing b^{-1/2}. These are not cosmetic typos because they change the fitted q. The paper reports one q value per dataset (1.703, 1.836, 2.106) with no error bars, no synthetic ground-truth tests, and code only 'available upon request.' Without a correct, validated CDF implementation, the observed disagreement with histogram methods cannot be interpreted as greater accuracy. I do not find the reader's sign-divergence objection to Eq. (9) to be the right diagnosis: the positive exponent in the denominator gives a convergent integral for ν>1/2; the real problem is that the integral is not the CDF and its constants are inconsistent.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper compares three estimators of the entropic index q for q-Gaussian distributions applied to three real-world time series: solar wind proton density fluctuations, Bitcoin log-returns, and Brazos River discharge. The three methods are direct nonlinear fitting to histogram PDFs, linear fitting via the q-logarithm, and a CDF-based fit. The main empirical claim is that histogram-based estimates of q vary substantially with the number of bins, whereas the CDF method gives one stable, bin-free estimate per dataset. The paper also compares q-Gaussian and Gaussian fits visually and reports that the three datasets are better described by q-Gaussians.","tokens_in":7589,"tokens_out":4306,"duration_ms":45293,"significance":"If the central claim were fully supported, the paper would make a useful methodological contribution: a bin-free estimation route for q-Gaussians is of practical interest across geophysics, space physics, and econophysics, and the bin-count sensitivity results in Figures 7-9 are a clear and instructive demonstration. The paper is also honest about its scope, explicitly not claiming a unified physical mechanism. However, the main result depends on an unverified and, as written, incorrect analytic expression for the CDF, and no synthetic ground-truth validation or uncertainty quantification is provided. These gaps currently prevent the stability claim from being established.","major_comments":[{"comment":"Equation (9) is not the CDF F(x)=∫_{-∞}^{x} p(x')dx' defined in Eq. (8). The integral shown runs over μ from 0 to ∞ and has no dependence on x, so it is the normalization integral over the half-line, not a cumulative distribution function. Consequently, the manuscript never specifies the theoretical curve actually used in the 'numerical fitting of the CDF'. Moreover, the constants in Eq. (9) are inconsistent: with p(x)=a(1+(b/ν)x^2)^{-ν}, the half-line integral is (a/2)(ν/b)^{1/2} B(1/2, ν−1/2), whereas Eq. (9) places a in the denominator and drops the factor ν^{1/2}. The positive exponent in the denominator is not by itself a divergence for ν>1/2; the load-bearing defect is that the expression is the wrong integral with incorrect prefactors. Because the fitted CDF q values (1.703, 1.836, 2.106) are the central evidence for the paper's main claim, this must be corrected with an explicit incomplete-beta expression for the CDF, and the fitted values must be recomputed and reported.","section":"Section 3.3, Eq. (9)"},{"comment":"The reported CDF q values are single point estimates with no error bars, confidence intervals, or bootstrap intervals, while the histogram-method scatter is shown only as raw dots. The text states that the first two methods 'begin to approach a fixed value, systematically higher' than the CDF value, but no quantitative comparison or uncertainty is given. Without an uncertainty measure for each q estimate, the claim that the CDF method is 'more stable' is not statistically supported; the apparent stability could be an artifact of a single deterministic fit, and the apparent disagreement with histogram methods could be within the (unreported) uncertainty of those methods.","section":"Section 4, Figures 7-9"},{"comment":"No simulation with a known q is performed, so the accuracy of any of the three estimators is not tested. The paper applies all methods to real data with unknown ground truth, which is sufficient to demonstrate bin sensitivity but not to conclude that the CDF method is more accurate or more reliable. A minimal validation on synthetic q-Gaussian samples with known q, covering the range q=1.5-2.5 and several sample sizes, would directly test which estimator recovers the true q and whether the CDF fit is indeed insensitive to binning. This is essential because the entire conclusion rests on the CDF implementation being correct.","section":"Sections 3-4, validation"},{"comment":"The code is only 'available upon request', and no version or execution details are given. For a paper whose central claim is a comparison of estimation algorithms, this is a serious reproducibility limitation: the reader cannot verify that the implemented CDF differs from the incorrect Eq. (9), nor reconstruct the exact preprocessing (e.g., the normalization procedure for dNp, the exact date ranges, and the handling of missing values). The authors should provide the code as supplementary material or a persistent repository.","section":"Section 6, Data availability"}],"minor_comments":[{"comment":"The y-axis label in Figure 2 reads 'Discharge[ft3/s]' but the figure shows Bitcoin closing prices, and the y-axis label in Figure 3 reads 'price(usd)' but the figure shows river discharge; the captions and axis labels are evidently swapped.","section":"Figures 2 and 3"},{"comment":"The sentence 'To ensure interdisciplinary rigor.' is an incomplete sentence and should be removed or merged with the following text.","section":"Section 2.1"},{"comment":"Section 5 states that solving the q-Gaussian integral 'requires the use of the incomplete beta function', but Section 3.3 only presents a complete beta function expression. The text should be consistent about which function is used in the numerical fit.","section":"Section 3.3"},{"comment":"Equation (2) is missing parentheses in the printed formula; it should read dNp(i) = 2*(Np(i+1)-Np(i))/(Np(i+1)+Np(i)). Also, in Section 2.2 the notations rt and Pb(i) are used without defining Pb explicitly as the Bitcoin closing price.","section":"Eq. (2) and Section 2.2"},{"comment":"There are typographical slips: 'linearaized' for 'linearized' in the figure captions, 'aksi' for 'also', and 'choose' for 'chosen'. These should be corrected in a careful copyedit.","section":"Figures 7-9 and text"},{"comment":"The caption says 'for 1980 to 2025', but the axis begins at 1984 and the text says '45 years' of data; the exact interval should be clarified.","section":"Figure 1 caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is not circular in the problematic sense: it compares estimators of the same quantity rather than fitting parameters and then calling them predictions. The main risk is purely technical: the CDF formula is wrong as printed, the implementation is unavailable, and there is no synthetic validation. If the authors correct the CDF expression, provide code, add uncertainty quantification, and run a synthetic data study, the revised paper could be a solid methodological contribution. I would not reject at this stage, but the central claim cannot be accepted on the current evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper does a useful, honest thing—shows on three real datasets how much histogram bin count moves q-Gaussian parameter estimates—but the CDF method that is supposed to replace histograms currently rests on an unverified numerical implementation, so the central \"more stable\" claim is not yet established.\n\nThe genuinely new part is the systematic side-by-side: solar wind proton density, Bitcoin returns, and Brazos River discharge, each fit with direct nonlinear PDF fit, q-log linearization, and CDF fit, with a clean sweep of bin counts from 10 to 200. Figures 7–9 tell a clear story: histogram-based q estimates wander by roughly 0.2–0.6 depending on bin count, while the CDF method outputs one number per dataset. That demonstration is worth having, and the authors are appropriately modest about not unifying mechanisms. The binning sensitivity itself is not new—the heavy-tail literature knows this—but the direct comparison across these three applied domains is a legitimate contribution.\n\nThe soft spot is load-bearing. Section 3.3 is supposed to provide the theoretical CDF used in the fit, but Eq. (9) is not F(x) at all; it is a half-line integral of the unnormalized kernel, i.e. a piece of the normalization constant. The stress-test note gets the diagnosis right: the printed expression is not divergent for q>1 with ν>1/2, but it is not the CDF, and the prefactor between the two equalities is inconsistent—a' order and powers of (b/ν) don't match. So a reader cannot reproduce the \"theoretical curve\" that was numerically integrated. Compounding that, no code is shipped, no synthetic-data test with known q is run, and no error bars appear on any of the reported q values. The phrase \"more stable\" at the end is honest—they say statistical tests are required—but stability is not accuracy, and without ground truth the systematic offset between CDF and histogram results (CDF always lower) might just be an artifact of a mis-normalized integrand.\n\nNone of this is fatal to the enterprise. The paper's positive message is that practitioners should stop trusting a single histogram fit. That advice stands even if the CDF implementation turns out to be correct after fixing. But the specific q values in the abstract (1.703, 1.836, 2.106) should be treated as provisional.\n\nThis is a paper for applied researchers who currently fit q-Gaussians to histograms and want a quick demonstration of the danger. It deserves a real peer-review pass, not a desk reject. I would send it back with a request for the actual incomplete-beta formula, the code, a synthetic validation, and error bars. If those come back, the paper could be solid. As it stands, it's a useful cautionary tale with a promising but unproven alternative.","headline":"A useful comparative warning about histogram bin-count sensitivity that is currently undermined by an unvalidated CDF implementation and no synthetic ground-truth checks.","tokens_in":8002,"tokens_out":2861,"would_cite":false,"duration_ms":29264,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Histogram binning destabilizes q-Gaussian fits; a bin-free CDF fit returns stable q values across solar wind, Bitcoin, and river data.","keywords":["q-Gaussian","entropic index q","heavy-tailed distributions","histogram binning","cumulative distribution function fitting","non-extensive statistics","solar wind fluctuations","financial returns"],"falsifier":"For each dataset, numerically integrate the $q$-Gaussian density $a(1-(1-q)b x^2)^{1/(1-q)}$ with the reported fitted $q$ and $b$ over the observed range and compare the result with the empirical CDF; if the residuals exceed those reported for the CDF fit, the CDF routine is not integrating the $q$-Gaussian.","tokens_in":6945,"feed_emoji":"📊","tokens_out":11857,"duration_ms":108693,"temperature":0.7,"pith_summary":"Three real-world fluctuation records—solar wind proton density, Bitcoin log-returns, and Brazos River daily discharge—are better described by heavy-tailed $q$-Gaussians than by normal distributions. The paper's central claim is that the two most common ways of estimating the entropic index $q$, direct nonlinear fits and $q$-logarithm linearization, both depend on an arbitrary histogram bin count and produce $q$ values that swing with that choice, whereas a numerical fit to the empirical cumulative distribution function (CDF) needs no bins and returns a single stable $q$ per system. For the three datasets the CDF method gives $q\\approx 1.703$, $1.836$, and $2.106$, respectively. A sympathetic reader should care because $q$ is used across geophysics, space weather, and finance to quantify heavy-tailed departure from Gaussian behavior, and the paper argues that this quantity cannot be trusted unless the fitting method itself is checked against the binning choice.","feed_headline":"Bin-free CDF fit stabilizes q-Gaussian estimates","feed_subtitle":"On solar wind, Bitcoin, and river discharge, the CDF fit returns one q while histogram fits shift with the bin count.","key_machinery":"The $q$-Gaussian density $Y(x)=a(1-(1-q)b x^2)^{1/(1-q)}$, which reduces to a Gaussian at $q=1$ and develops power-law tails for $q>1$. The paper compares three estimation devices: direct nonlinear least-squares fitting of this density to normalized histogram heights; iterative linearization through the $q$-logarithm $\\ln_q Y = (Y^{1-q}-1)/(1-q)$, testing trial $q$ values from 1 to 3 in steps of 0.01 and selecting the best correlation; and a nonlinear fit of the empirical CDF to a numerically integrated theoretical CDF, with the integral expressed through the Beta function. The CDF fit is the load-bearing mechanism because it removes the binning step and is the only method that returns one $q$ per dataset rather than a curve of $q$ versus bin count.","core_discovery":"The paper establishes that the empirical distributions of solar wind proton density fluctuations, Bitcoin log-returns, and Brazos River discharge are all consistent with $q$-Gaussian heavy-tailed statistics, and that the estimated $q$ is not a property of the data alone: with histogram-based fits it moves systematically with the number of bins, while the cumulative distribution function fit yields a bin-free estimate. The reported values are $q\\approx 1.703$ (solar wind), $q\\approx 1.836$ (Bitcoin), and $q\\approx 2.106$ (river discharge). Because the histogram methods converge to values systematically above the CDF value at high bin counts, the paper concludes that CDF fitting gives more stable estimates and that binning is a critical methodological choice in $q$ determination.","pith_inferences":["Editorial note: the Beta-function expression printed as Eq. (9) carries a positive exponent $[a(1+b\\mu/\\nu)]^\\nu$, whereas for $q>1$ the $q$-Gaussian integrand is $(1+(q-1)b x^2)^{-1/(q-1)}$; if the implemented code follows the printed formula, the reported CDF values are not derived from the fitted density.","If the CDF method is verified by direct numerical integration, the systematic gap between histogram and CDF $q$ values suggests a recalibration exercise: apply both methods to archived datasets to estimate a correction for existing histogram-based $q$ values.","The same bin-free principle can be tested on other heavy-tailed families, such as $q$-exponentials or kappa distributions, by comparing CDF fits to histogram fits on synthetic data with known parameters."],"forward_implications":["For the three analyzed systems, histogram-based $q$ estimates should be reported with a bin-count sensitivity scan; across 10 to 200 bins the fitted $q$ varies by more than a unit in the river-discharge case.","A single bin-free $q$ per system is available from the CDF method: $q\\approx 1.703$ for solar wind proton density fluctuations, $q\\approx 1.836$ for Bitcoin log-returns, and $q\\approx 2.106$ for Brazos River discharge.","The CDF method can be applied without modification to other heavy-tailed records, since it eliminates the bin-width choice.","Because high-bin histogram fits converge to values systematically higher than the CDF value, earlier histogram-based $q$ estimates are likely to have overstated the degree of heavy-tailed behavior."],"supporting_citations":[{"why":"Introduces the non-extensive entropy whose extremization gives the q-Gaussian family being fitted.","marker":"Tsallis, 1988"},{"why":"Derives the q-Gaussian form and the entropic index parametrization used throughout the comparison.","marker":"Tsallis et al., 2005"},{"why":"Documents the bias that histogram binning introduces in heavy-tailed estimation, motivating the bin-free CDF comparison.","marker":"Nair et al., 2022"},{"why":"Supplies the hourly solar wind proton density data used for the first system.","marker":"King and Papitashvili, 2005"},{"why":"Provides evidence that solar wind fluctuations follow heavy-tailed distributions, justifying the solar wind case.","marker":"Burlaga and Ness, 2009"},{"why":"Establishes prior q-based analysis of cryptocurrency returns, supporting Bitcoin as a heavy-tailed financial system.","marker":"Stosic et al., 2018a"},{"why":"Establishes prior q-triplet analysis of Brazos River discharge, supporting the hydrological system's heavy-tailed behavior.","marker":"Stosic et al., 2018b"},{"why":"Provides the q-central limit theorem giving theoretical grounds for expecting q-Gaussian fluctuations.","marker":"Umarov et al., 2008"}],"fun_headline_variants":["Bin-free CDF fit stabilizes q-Gaussian estimates","CDF fitting avoids binning bias in q-Gaussian analyses","Histogram binning shifts q values; CDF fit offers stable alternative","CDF method gives robust q for heavy-tailed data","CDF fit yields stable q across solar wind, Bitcoin, and rivers"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's conclusion stands on the assumption that its numerical CDF routine correctly evaluates the $q$-Gaussian cumulative distribution; the printed analytic expression in Eq. (9) diverges for $q>1$ and the source code is not supplied.","fun_headline_variants_meta":{"raw":{"variants":["Bin-free CDF fit stabilizes q-Gaussian estimates","CDF fitting avoids binning bias in q-Gaussian analyses","Histogram binning shifts q values; CDF fit offers stable alternative","CDF method gives robust q for heavy-tailed data","CDF fit yields stable q across solar wind, Bitcoin, and rivers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000692,"raw_usage":{"total_tokens":3110,"prompt_tokens":901,"completion_tokens":2209,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":2132}},"tokens_in":517,"tokens_out":2209,"duration_ms":15388,"temperature":1.0,"reasoning_tokens":2132,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:26:27.145589+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For each dataset, numerically integrate the $q$-Gaussian density $a(1-(1-q)b x^2)^{1/(1-q)}$ with the reported fitted $q$ and $b$ over the observed range and compare the result with the empirical CDF; if the residuals exceed those reported for the CDF fit, the CDF routine is not integrating the $q$-Gaussian.","supporting_citations":[],"review_version":1}