{"id":"dfbd2ba6-7d84-46d9-bb9c-89833257710d","arxiv_id":"2508.08934","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":1,"one_line_summary":"Using the FGM copula, the authors derive analytical first-order corrections due to consecutive-point correlation for HVG and DHVG degree distributions.","lead":"This paper derives mathematical formulas for the degree distributions of horizontal visibility graphs when successive data points in a time series are correlated. The work offers a rigorous route to interpret network measures of correlated time series.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim likely under-specified: a bivariate FGM copula for consecutive pairs does not determine the multivariate block distributions needed for HVG degrees.","rationale":"The reader's verdict was UNVERDICTED because only the abstract was available. My concern is distinct from the reader's weakest assumption: the reader emphasized that FGM captures only weak, linear-order dependence and may be unrealistic for real time series. I focus instead on whether the mathematical object is well-defined at all, given that a bivariate copula for consecutive pairs does not fix the multivariate block distributions that HVG degrees depend on. This is an internal-consistency concern rather than a realism concern. Since the full text is unavailable, I cannot determine whether the paper already resolves this by assuming, say, a Markovian structure or a multivariate FGM copula with vanishing higher-order interactions. Therefore the appropriate verdict remains UNVERDICTED, and the reader's verdict need not change. The proposed concrete test would settle the concern by checking whether the derivation explicitly specifies a consistent joint distribution for blocks of length >= 3 and whether the degree-distribution formulas are invariant to alternative extensions with the same bivariate margins.","tokens_in":695,"tokens_out":4256,"duration_ms":52125,"concrete_test":"Inspect the full derivation and identify the exact joint model used for blocks of length m >= 3. Specifically, compute P(X_1 > X_2, X_2 < X_3) under the paper's model. If the model is defined only by bivariate FGM copulas on consecutive pairs, this probability is not uniquely determined. Compare two natural extensions: (a) a Markov chain with FGM transition copula, and (b) a trivariate FGM copula with theta_12 = theta_23 = theta and theta_13 = 0. If the resulting HVG degree distributions differ, the paper must explicitly commit to one extension; otherwise the central claim is ambiguous and the formulas cannot be uniquely verified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The horizontal visibility graph degree of a node depends on joint events over blocks of consecutive observations. For example, visibility between nodes i and i+m requires every intermediate point to lie below the line segment, an event governed by the joint distribution of (X_i, X_{i+1}, ..., X_{i+m}). The abstract states that the FGM copula is used to model correlation between consecutive data points, but a bivariate FGM copula specifies only lag-1 pairwise dependence. It does not uniquely determine the joint distribution of triples, quadruples, or larger blocks. Unless the paper explicitly assumes a Markov chain with FGM transition copula, or a multivariate FGM copula with specified higher-order interaction terms, the derived degree distributions are not well-defined functions of the stated input. The phrase 'correlation between consecutive data points' suggests only bivariate information, which is insufficient for a graph-theoretic quantity that depends on longer-range order statistics. Thus the central claim, as stated, risks being underdetermined: different consistent extensions of the bivariate FGM model could yield different HVG degree distributions, so the claimed analytical result may describe one particular model rather than 'correlated time series' generally.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims to derive analytical expressions for the degree distributions of horizontal visibility graphs (HVGs) and directed horizontal visibility graphs (DHVGs) for correlated time series. The method uses the Farlie-Gumbel-Morgenstern (FGM) copula to model the dependence between consecutive data points, and the reported results are said to be exact up to first order in the FGM copula parameter θ. The abstract presents this as a step toward rigorous analytical understanding of visibility graphs beyond the i.i.d. case.","tokens_in":922,"tokens_out":2734,"duration_ms":30230,"significance":"If correct, this is a potentially valuable contribution: exact, first-order-in-θ degree distributions for HVGs and DHVGs would give the visibility-graph community one of the first analytical handles on correlated time series, complementing the existing exact results for uncorrelated series. The claim is specific and falsifiable, and the FGM copula provides a tractable weak-dependence model. However, the significance is conditional on the model being fully specified and on the derivations being correct; neither can be verified from the abstract alone.","major_comments":[{"comment":"The central claim is under-specified regarding the probabilistic model. The HVG degree of a node is determined by joint visibility events over blocks of observations; for example, visibility between nodes i and i+m requires every intermediate observation to lie below the line segment, which depends on the joint distribution of (X_i, X_{i+1}, ..., X_{i+m}). The abstract mentions only that the FGM copula models 'correlation between consecutive data points,' which reads as a bivariate lag-1 specification. A bivariate FGM copula does not uniquely determine the joint distribution of triples, quadruples, or larger blocks, so the derived degree distributions would not be well-defined functions of the stated input unless the paper explicitly assumes a multivariate FGM copula with specified higher-order interactions, or a Markov chain with an FGM transition copula, or an equivalent construction. The authors must state the full model; otherwise the claimed analytical results describe one implicit model rather than 'correlated time series' generally.","section":"Abstract"},{"comment":"The claim of exactness 'up to the first order of the correlation parameter' requires a proof that higher-order terms in θ (order θ^2 and beyond) can be uniformly controlled or discarded for all block lengths that contribute to the degree distribution. The abstract provides no indication of how this is established, and it is not a trivial property: the number of contributing blocks grows with the distance between nodes, so the accumulation of higher-order terms must be shown to remain subleading. Without seeing the derivation, this is a load-bearing point that cannot be checked from the abstract alone.","section":"Abstract"}],"minor_comments":[{"comment":"The abstract uses the notation HVGs and DHVGs without defining the directed version; a brief parenthetical definition would help readers unfamiliar with the variant.","section":"Abstract"},{"comment":"The phrase 'adopt the FGM copula method' could be made more precise by indicating whether the copula is used only for the bivariate lag-1 dependence or for the entire multivariate joint distribution; this would reduce ambiguity in the central claim.","section":"Abstract"}],"recommendation":"uncertain","confidential_remarks":"This is an abstract-only review. The soundness of the derivations cannot be evaluated, and the main conceptual concern—the underdetermination of block distributions from a bivariate FGM model—needs to be resolved by the full text. If the full paper specifies a proper multivariate construction and includes the detailed first-order calculations, I would be willing to re-review; as it stands, the evidence is insufficient for a soundness judgment."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I only have the abstract, so this is a provisional read, not a verdict. But here is what I would want you to know before you spend time on this: the claimed result is precisely the kind of gap the visibility-graph field has been circling. For all the practical use of VGs, analytical results for correlated time series are genuinely scarce, and an exact first-order-in-θ expression for HVG and DHVG degree distributions would be a useful, citable contribution if the derivation holds.\n\nWhat looks good from the abstract: the authors use the FGM copula, which is simple enough to keep calculations tractable, and they explicitly report the dependence on the correlation parameter. That is a sensible way to get analytical leverage, and the claim of exactness to first order is concrete and falsifiable. If they actually deliver that, it fills a recognized gap.\n\nNow the soft spots, in proportion. The stress-test note you passed along raises a real mathematical concern: HVG degrees depend on joint events over blocks of consecutive points, not just on pairwise lag-1 dependence. A bivariate FGM copula specifies only pairwise structure. Unless the paper assumes a Markov chain with FGM transitions, or a multivariate FGM with specified higher-order terms, the degree distribution may not be uniquely defined from the stated input. That is not a trivial technicality; it goes to whether the formulas describe a well-defined model or a class of models. The abstract does not say enough to rule the concern in or out.\n\nSecond, even if the multivariate extension is consistent, FGM copulas capture only weak dependence. The correlation parameter θ is bounded in a narrow range, so the practical range of serial correlation covered may be small. That is not a flaw, but it tempers the title's claim about \"correlated time series.\"\n\nMy take: this is a paper worth reading in full. The math needs checking, and the multivariate consistency issue is the first place to look. If the derivation is correct, it deserves publication in a good applied-math or network-science venue. I would send it to a serious referee rather than desk-reject it. For my own work, I would not cite it until I see the full derivation.\n\nRecommendation: engage with the full text, but treat the bivariate-to-multivariate step as the load-bearing point.","headline":"Abstract-only, but the claimed first-order analytical HVG results are exactly what the visibility-graph literature lacks; the key question is whether the FGM model actually determines the block distributions the derivation needs.","tokens_in":1380,"tokens_out":1616,"would_cite":false,"duration_ms":19864,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["62H05","62M10","05C80"],"pacs":[],"model":"deepseek-v4-flash","headline":"For correlated time series, the degree distributions of horizontal and directed horizontal visibility graphs are derived exactly to first order in the FGM copula correlation parameter.","keywords":["horizontal visibility graph","degree distribution","Farlie-Gumbel-Morgenstern copula","correlated time series","directed horizontal visibility graph","analytical solution","first-order approximation"],"falsifier":"Simulate a time series whose consecutive pairs are drawn from an FGM copula with known $\\theta$, build its horizontal visibility graph, and compare the empirical degree distribution with the analytic first-order prediction. Then repeat with a non-FGM copula (e.g., a Gaussian copula) having the same linear correlation: if the deviations from the first-order formula do not remain at second order in $\\theta$—or if the FGM prediction already mismatches at first order—the assumption fails.","tokens_in":530,"feed_emoji":"📈","tokens_out":4087,"duration_ms":40967,"temperature":0.7,"pith_summary":"This paper aims to close a gap in the theory of visibility graphs: for correlated time series, no rigorous analytical formulas for degree distributions existed. The authors adopt the Farlie-Gumbel-Morgenstern (FGM) copula to model the dependence between consecutive data points and derive exact expressions for the degree distributions of the horizontal visibility graph (HVG) and its directed version (DHVG) up to first order in the copula parameter. If correct, these results give a tractable, parameter-explicit link between serial correlation and network structure, allowing predictions of how correlation reshapes visibility graphs without simulations.","feed_headline":"First-order formula links correlation to visibility graph degrees","feed_subtitle":"A copula-based derivation gives exact degree distributions for horizontal visibility graphs of correlated series.","key_machinery":"The Farlie-Gumbel-Morgenstern (FGM) copula is the central object: a one-parameter family of bivariate distributions whose joint density is a linear perturbation of independence, controlled by the correlation parameter $\\theta$. Written into the joint distribution of consecutive data points, it makes the geometric horizontal visibility condition—where an intervening point blocks a line of sight between two values—amenable to closed-form averaging. The copula supplies a tractable correlation structure, and the first-order expansion in $\\theta$ carries the entire argument from the copula to the degree distribution.","core_discovery":"The paper claims that when the joint distribution of consecutive time series values is modeled by an FGM copula with parameter $\\theta$, the connection probability between two nodes in an HVG—and hence the resulting degree distribution—can be computed exactly to first order in $\\theta$. The same holds for the directed HVG, where in- and out-degree distributions are obtained separately. Concretely, the degree distribution splits into the independent case ($\\theta=0$) plus a correction term linear in $\\theta$, so the effect of correlation is understood at leading order. This provides a first rigorous, analytical handle on how serial dependence alone modifies the network representation of a time series.","pith_inferences":["If $\\theta$ is translated into a lag-1 linear correlation coefficient, the derived formulas predict HVG degree shifts proportional to that coefficient; this could be tested on simulated Gaussian-copula or AR(1) series to map the regime where FGM is an adequate model.","The same FGM machinery might extend to other visibility graph variants (e.g., parametric visibility graphs) or to non-consecutive joint distributions, though the paper does not claim these extensions.","Because FGM captures only weak, linear-order dependence, the leading-order results serve as a minimal benchmark: real long-range correlated data should show deviations beyond first order, and those deviations could be used to detect structure that a linear copula misses.","The first-order correction term is likely a computable sum over visibility configurations; extracting that sum could reveal combinatorial reasons why particular degree sequences are favored under correlation."],"forward_implications":["For a series whose consecutive dependence is FGM, the HVG and DHVG degree distributions can now be computed analytically to first order in the correlation strength without Monte Carlo simulation.","Positive versus negative serial correlation is predicted to shift the degree distribution in opposite directions, offering a qualitative network-level fingerprint of the sign of dependence.","The $\\theta=0$ limit must reproduce the known i.i.d. degree distributions, providing a built-in consistency check for the derivation.","The directed version yields separate in- and out-degree distributions, so asymmetries in serial dependence become visible in directed network statistics.","The explicit formulas open the way to estimating the FGM correlation parameter from observed HVG degree sequences by fitting the first-order correction."],"supporting_citations":[],"fun_headline_variants":["Copula math cracks exact degree laws for correlated HVG","Exact first-order degree distributions for correlated visibility graphs","Serial correlation's exact effect on visibility graph degrees","FGM copula yields precise degree formulas for HVG"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation assumes that the dependence between consecutive data points is exactly described by an FGM copula with a single parameter $\\theta$, which only captures weak, linear-order dependence; if the real serial dependence is stronger or of a different functional form, the first-order formulas will not describe the actual HVG degree distribution.","fun_headline_variants_meta":{"raw":{"variants":["Copula math cracks exact degree laws for correlated HVG","Exact first-order degree distributions for correlated visibility graphs","Serial correlation's exact effect on visibility graph degrees","FGM copula yields precise degree formulas for HVG"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001543,"raw_usage":{"total_tokens":6123,"prompt_tokens":846,"completion_tokens":5277,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":462,"completion_tokens_details":{"reasoning_tokens":5213}},"tokens_in":462,"tokens_out":5277,"duration_ms":36244,"temperature":1.0,"reasoning_tokens":5213,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:30:59.417400+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate a time series whose consecutive pairs are drawn from an FGM copula with known $\\theta$, build its horizontal visibility graph, and compare the empirical degree distribution with the analytic first-order prediction. Then repeat with a non-FGM copula (e.g., a Gaussian copula) having the same linear correlation: if the deviations from the first-order formula do not remain at second order in $\\theta$—or if the FGM prediction already mismatches at first order—the assumption fails.","supporting_citations":[],"review_version":2}