{"paper_id":"2305.12632","title":"Deformation of Marchenko-Pastur distribution for the correlated time series","authors":["Masato Hisakado","Takuya Kaneko"],"paper_date":"2023-05-22","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2305.12632v2","page_url":"https://thequant.space/flowcharts/deformation-of-marchenko-pastur-distribution-for-the-correla/","scores":{"empirical_rigor":3,"math_complexity":8.5,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes to study how temporal correlations in time series deform the Marchenko‑Pastur eigenvalue distribution of the Wishart matrix, analyzing moments and phase‑transition behavior.","quote":"We study the eigenvalue of the Wishart matrix, which is created from a time series with temporal correlation. When there is temporal correlation, the eigenvalue distribution converges to the deformed MPD which has a longer tail and higher peak than the MPD.","summary":"We study the eigenvalue of the Wishart matrix, which is created from a time series with temporal correlation. When there is no correlation, the eigenvalue distribution of the Wishart matrix is known as the Marchenko-Pastur distribution (MPD) in the double scaling limit. When there is temporal…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The time series are normalized with zero mean and unit variance.","type":"model","stated":"explicit","quote":"Here we set the normalization, E(At) = 0, and V(At) = 1.","support":"full"},{"assumption":"Temporal correlation functions follow either exponential decay or power‑law decay.","type":"model","stated":"explicit","quote":"The temporal correlation is the exponential decay and power decay. In this article we consider the exponential decay and power decay cases.","support":"full"},{"assumption":"In the large‑size limit, correlations beyond lag L vanish (d_L ≈ 0).","type":"statistical","stated":"explicit","quote":"Second moment µ2 = 1 L2N L X ν1=1 L X ν2=1 N X m1=1 N X m2=1 < AT m1ν1Aν1m2AT m2ν2Aν2m1 > = 1 L2N L X ν1=1 L X ν2=1 N X m1=1 N X m2=1 < Aν1m1Aν1m2Aν2m2Aν2m1 >= 1 + 1 Q + 2 Q X i=1 d2 i , (11) in the limit of N, L →∞with L/N = Q. Here we assume dL ∼0 in this limit, because < Aj, Aj+L >= 0, in the limit L →∞.","support":"full"},{"assumption":"The Wishart matrix is constructed from a single product’s time series folded into an L×N matrix.","type":"data","stated":"explicit","quote":"Here we fold the time series At, N times, where T = N × L, A = (A1, A1+L, · · · , A1+(N−1)L), (3) where Aµ = (Aµ, · · · , Aµ+L−1)T, µ = 1, 1 + L, · · · , 1 + (N −1)L, is the size L vector and XT is the transpose of the vector X. Aµ is the time series from µ to µ + L −1. In the application of the random matrix to the finance, Aµ usually corresponds to the different product time series [5, 6], but here we use one","support":"full"},{"assumption":"The double‑scaling limit (N, L → ∞ with L/N = Q) is taken for analytical results.","type":"model","stated":"explicit","quote":"First moment µ1 = 1 LN L X ν=1 N X m=1 < (Aνm)2 >= 1, (10) in the limit of N, L →∞with L/N = Q. Note that it does not depend on the type of correlation decay. Second moment µ2 = 1 L2N L X ν1=1 L X ν2=1 N X m1=1 N X m2=1 < AT m1ν1Aν1m2AT m2ν2Aν2m1 > = 1 L2N L X ν1=1 L X ν2=1 N X m1=1 N X m2=1 < Aν1m1Aν1m2Aν2m2Aν2m1 >= 1 + 1 Q + 2 Q X i=1 d2 i , (11) in the limit of N, L →∞with L/N = Q. Here we assume dL ∼0 in this","support":"full"},{"assumption":"Temporal correlation parameter d_i is small enough that its contribution can be treated as a perturbation in the second‑moment formula.","type":"statistical","stated":"implied","quote":"It is 1 L2N X ν1̸=ν2 X m1̸=m2 < Aν1m1Aν1m2Aν2m2Aν2m1 >= N(N −1) L2N X |ν1−ν2| < Aν1m1Aν2m1 >2 7 = 2N(N −1)L L2N X i d2 i = 2 Q X i=1 d2 i . (12) The second moment of Eq.(11) increases as the temporal correlation increases.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"USD/CAD","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"Data µ2 µ3 µ4 µ5 µ6 MPD 1.333 2.111 3.704 6.938 13.597 1 USD/CAD 1.419 (0.008) 2.448 (0.028) 4.716 (0.082) 9.732 (0.218) 21.038 (0.568) 2 EUR/CHF 1.430 (0.003) 2.494 (0.016) 4.856 (0.055) 10.113 (0.164) 22.032 (0.456) 3 EUR/GBP 1.402 (0.003) 2.385 (0.015) 4.533 (0.055) 9.237 (0.180) 19.738 (0.562) 4 SOY 1.333 (0.003) 2.111 (0.013) 3.703 (0.041) 6.933 (0.122) 13.569 (0.344) 5 VIX 1.336 (0.007) 2.117 (0.027) 3.713","support":"full"},{"name":"EUR/CHF","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"Data µ2 µ3 µ4 µ5 µ6 MPD 1.333 2.111 3.704 6.938 13.597 1 USD/CAD 1.419 (0.008) 2.448 (0.028) 4.716 (0.082) 9.732 (0.218) 21.038 (0.568) 2 EUR/CHF 1.430 (0.003) 2.494 (0.016) 4.856 (0.055) 10.113 (0.164) 22.032 (0.456) 3 EUR/GBP 1.402 (0.003) 2.385 (0.015) 4.533 (0.055) 9.237 (0.180) 19.738 (0.562) 4 SOY 1.333 (0.003) 2.111 (0.013) 3.703 (0.041) 6.933 (0.122) 13.569 (0.344) 5 VIX 1.336 (0.007) 2.117 (0.027) 3.713","support":"full"},{"name":"EUR/GBP","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"Data µ2 µ3 µ4 µ5 µ6 MPD 1.333 2.111 3.704 6.938 13.597 1 USD/CAD 1.419 (0.008) 2.448 (0.028) 4.716 (0.082) 9.732 (0.218) 21.038 (0.568) 2 EUR/CHF 1.430 (0.003) 2.494 (0.016) 4.856 (0.055) 10.113 (0.164) 22.032 (0.456) 3 EUR/GBP 1.402 (0.003) 2.385 (0.015) 4.533 (0.055) 9.237 (0.180) 19.738 (0.562) 4 SOY 1.333 (0.003) 2.111 (0.013) 3.703 (0.041) 6.933 (0.122) 13.569 (0.344) 5 VIX 1.336 (0.007) 2.117 (0.027) 3.713","support":"full"},{"name":"SOY","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"Data µ2 µ3 µ4 µ5 µ6 MPD 1.333 2.111 3.704 6.938 13.597 1 USD/CAD 1.419 (0.008) 2.448 (0.028) 4.716 (0.082) 9.732 (0.218) 21.038 (0.568) 2 EUR/CHF 1.430 (0.003) 2.494 (0.016) 4.856 (0.055) 10.113 (0.164) 22.032 (0.456) 3 EUR/GBP 1.402 (0.003) 2.385 (0.015) 4.533 (0.055) 9.237 (0.180) 19.738 (0.562) 4 SOY 1.333 (0.003) 2.111 (0.013) 3.703 (0.041) 6.933 (0.122) 13.569 (0.344) 5 VIX 1.336 (0.007) 2.117 (0.027) 3.713","support":"full"},{"name":"VIX","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"Data µ2 µ3 µ4 µ5 µ6 MPD 1.333 2.111 3.704 6.938 13.597 1 USD/CAD 1.419 (0.008) 2.448 (0.028) 4.716 (0.082) 9.732 (0.218) 21.038 (0.568) 2 EUR/CHF 1.430 (0.003) 2.494 (0.016) 4.856 (0.055) 10.113 (0.164) 22.032 (0.456) 3 EUR/GBP 1.402 (0.003) 2.385 (0.015) 4.533 (0.055) 9.237 (0.180) 19.738 (0.562) 4 SOY 1.333 (0.003) 2.111 (0.013) 3.703 (0.041) 6.933 (0.122) 13.569 (0.344) 5 VIX 1.336 (0.007) 2.117 (0.027) 3.713","support":"full"},{"name":"NKY 225","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"(0.736) 6 NKY 225 1.324 (0.049) 2.081 (0.020) 3.637 (0.064) 6.819 (0.186) 13.433 (0.512) TABLE II.","support":"full"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"unclear"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"The second moment of the eigenvalue distribution increases as temporal correlation increases.","kind":"statistical","quote":"We show that the second moment increases as the temporal correlation increases. We show that the mean of the deformed MPD does not depend on the correlation and that the second moment increases as the temporal correlation increases.","support":"full"},{"finding":"The deformed Marchenko‑Pastur distribution has a longer tail and higher peak compared with the standard MPD.","kind":"theoretical","quote":"When there is temporal correlation, the eigenvalue distribution converges to the deformed MPD which has a longer tail and higher peak than the MPD. Hence, as the correlation increases, the distribution has a fatter tail and a higher peak.","support":"full"},{"finding":"A phase transition occurs at power‑law index γ = 1/2: for γ > 1/2 the second moment and largest eigenvalue are finite, while for γ ≤ 1/2 they diverge.","kind":"theoretical","quote":"When $γ>1/2$ which is the power index of the temporal correlation, the second moment of the distribution is finite and the largest eigenvalue is finite. On the other hand, when $γ\\leq 1/2$, the second moment is infinite and the largest eigenvalue is infinite.","support":"full"},{"finding":"Finite‑size scaling analysis yields a critical exponent β = 1 for the phase transition.","kind":"theoretical","quote":"Using finite scaling analysis, we estimate the critical exponent of the phase transition. Hence, we can estimate the critical exponent, m ∝tβ at H ∼Hc (24) where β = 1 and ξ∞∝t−ν = t−4/3 at H ∼Hc, (25) where ξ∞is the correlation length. VI.","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":16,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":2,"extracted_at":"2026-11-14T20:02:50Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}