pith. sign in

arxiv: 2504.21647 · v3 · submitted 2025-04-30 · 📊 stat.ME · math.ST· stat.ML· stat.TH

Conditional independence testing with a single realization of a multivariate nonstationary nonlinear time series

Michael Wieck-Sosa , Michel F. C. Haddad , Aaditya Ramdas This is my paper

Pith reviewed 2026-05-22 18:00 UTC · model grok-4.3

classification 📊 stat.ME math.STstat.MLstat.TH
keywords conditional independence testingnonstationary time seriesnonlinear processessingle realizationcausal discoverylong-run covarianceGaussian approximationmultivariate time series
0
0 comments X

The pith

A framework enables conditional independence testing from a single realization of a nonstationary nonlinear multivariate time series.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a method to test conditional independence among components of a multivariate time series when only one observed path is available and the underlying process is allowed to be both nonstationary and nonlinear. Existing tools for time-series conditional independence either require multiple independent realizations or impose linearity and stationarity, which excludes many practical datasets such as single economic trajectories or climate records. The new procedure first fits time-varying nonlinear regressions to isolate error processes, then builds test statistics from local long-run covariances of products of those errors, and finally uses a distribution-uniform strong Gaussian approximation to obtain critical values. If the procedure is valid, it supplies p-values that can be used directly for causal discovery and variable selection in previously inaccessible single-trajectory settings.

Core claim

The central claim is that conditional independence between two components of a multivariate time series, given a third, can be tested using only a single realization even when the series is nonstationary and the dependence structure is nonlinear. The test is constructed by estimating time-varying nonlinear regression functions to produce error processes, forming statistics from local long-run covariance matrices of products of these errors, and applying a distribution-uniform strong Gaussian approximation to control the test.

What carries the argument

Time-varying nonlinear regression combined with estimation of local long-run covariance matrices of error products and a distribution-uniform strong Gaussian approximation.

If this is right

  • Tests for unconditional independence between two processes become available under the same single-realization nonstationary nonlinear conditions.
  • Causal discovery algorithms that rely on conditional independence tests can now be applied to single observed trajectories of nonlinear nonstationary series.
  • Variable selection procedures for multivariate time series can operate without requiring multiple independent realizations or linearity.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same machinery could be used to test conditional independence in single long recordings from neuroscience or finance where stationarity cannot be assumed.
  • If the Gaussian approximation holds uniformly, the framework might extend to constructing confidence intervals for measures of nonlinear dependence in nonstationary settings.

Load-bearing premise

The method assumes that time-varying nonlinear regression functions can be estimated reliably enough from a single realization to yield error processes whose local long-run covariances admit a consistent Gaussian approximation.

What would settle it

A simulation experiment on a known nonstationary nonlinear process in which conditional independence holds but the proposed test rejects at rates far above the nominal level across many sample sizes would falsify the claim.

Figures

Figures reproduced from arXiv: 2504.21647 by Aaditya Ramdas, Michael Wieck-Sosa, Michel F. C. Haddad.

Figure 1
Figure 1. Figure 1: Causal graph depicting the time-delayed causal relationships among the stochastic [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Our test holds the level even with fairly small sample sizes, and gains power as [PITH_FULL_IMAGE:figures/full_fig_p031_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: One realization from the null distribution with [PITH_FULL_IMAGE:figures/full_fig_p031_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: One realization from the alternative distribution with [PITH_FULL_IMAGE:figures/full_fig_p031_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The time-varying regression function fK(z, u) from (14) at different u and K. -0.25 0.00 0.25 0.50 -2 0 2 z gK(z, u) u = 0.0, K = 1 -0.25 0.00 0.25 0.50 -2 0 2 z gK(z, u) u = 0.5, K = 1 -0.25 0.00 0.25 0.50 -2 0 2 z gK(z, u) u = 0.0, K = 4 -0.25 0.00 0.25 0.50 -2 0 2 z gK(z, u) u = 0.5, K = 4 [PITH_FULL_IMAGE:figures/full_fig_p032_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The time-varying regression function gK(z, u) from (15) at different u and K. 5 Real Data Application We investigate how stock markets in the United States, United Kingdom, Hong Kong, and Japan are linked. The dataset consists of daily log returns based on the adjusted closing prices of the S&P 500, FTSE 100, Hang Seng, and Nikkei 225 from January 2022 to March 2025. To deal with holidays observed by each … view at source ↗
Figure 7
Figure 7. Figure 7: In this setup with identical time-varying regression functions, the Sieve-dGCM test [PITH_FULL_IMAGE:figures/full_fig_p095_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Our test holds the level even with fairly small sample sizes, and gains power as [PITH_FULL_IMAGE:figures/full_fig_p106_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: One realization from the null distribution with [PITH_FULL_IMAGE:figures/full_fig_p106_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: One realization from the alternative distribution with [PITH_FULL_IMAGE:figures/full_fig_p107_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: The time-varying mean function µ X Ψ (u) from (27) at different complexities. 0.2 0.3 0.4 0.00 0.25 0.50 0.75 1.00 u μY Ψ(u) Ψ = 1 0.2 0.3 0.4 0.00 0.25 0.50 0.75 1.00 u μY Ψ(u) Ψ = 4 [PITH_FULL_IMAGE:figures/full_fig_p107_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: The time-varying mean function µ Y Ψ(u) from (28) at different complexities. D.6 Alternative Test Statistics Consider the test statistic S ⋆ n,p(Rˆn) = [PITH_FULL_IMAGE:figures/full_fig_p107_12.png] view at source ↗
read the original abstract

Identifying relationships among stochastic processes is a core objective in many fields, such as economics. While the standard toolkit for multivariate time series analysis has many advantages, it can be difficult to capture nonlinear dynamics using linear vector autoregressive models. This difficulty has motivated the development of methods for causal discovery and variable selection for nonlinear time series, which routinely employ tests for conditional independence. In this paper, we introduce the first framework for conditional independence testing that works with a single realization of a nonstationary nonlinear process. We also show how our framework can be used to test for independence. The key technical ingredients of our framework are time-varying nonlinear regression, estimation of local long-run covariance matrices of products of error processes, and a distribution-uniform strong Gaussian approximation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper introduces the first framework for conditional independence testing (and independence testing) that operates on a single realization of a multivariate nonstationary nonlinear time series. The method combines time-varying nonlinear regression, local long-run covariance estimation of products of error processes, and a distribution-uniform strong Gaussian approximation to construct valid test statistics and critical values.

Significance. If the technical claims hold, the result would fill an important gap by extending conditional independence testing beyond the stationary or multi-realization settings that dominate the literature. This is relevant for empirical work in economics and related fields where only one observed path of a nonlinear nonstationary process is available. The paper explicitly builds on standard regression and Gaussian-approximation tools rather than introducing entirely new primitives, which strengthens its potential impact if the uniformity conditions are rigorously established.

major comments (1)
  1. [§3.2–3.3, Theorem 4.1] §3.2–3.3 and Theorem 4.1: The distribution-uniform strong Gaussian approximation for the local long-run covariance estimators of residual products is presented as holding under the single-realization nonstationary nonlinear regime. However, the effective local sample size is governed by the bandwidth, which shrinks with sample size, while the regression function itself is time-varying; the uniformity over the function class therefore rests on moment and dependence conditions whose verification is not fully detailed for this regime. Because the critical values for the conditional-independence test are obtained directly from this approximation, any gap here is load-bearing for the central claim.
minor comments (2)
  1. [Notation] Notation for the time-varying regression estimator and the local covariance estimator should be made consistent between the main text and the appendix; currently the same symbol is used for both the population and estimated versions in several places.
  2. [Simulation study] The simulation section would benefit from an explicit statement of the bandwidth selection rule used in the reported experiments, as this choice directly affects the local sample size and therefore the finite-sample behavior of the Gaussian approximation.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful and constructive report. The concern about the distribution-uniform strong Gaussian approximation is well taken, as it underpins the validity of the critical values. We respond to the major comment below and will strengthen the exposition in the revision.

read point-by-point responses

  1. Referee: [§3.2–3.3, Theorem 4.1] §3.2–3.3 and Theorem 4.1: The distribution-uniform strong Gaussian approximation for the local long-run covariance estimators of residual products is presented as holding under the single-realization nonstationary nonlinear regime. However, the effective local sample size is governed by the bandwidth, which shrinks with sample size, while the regression function itself is time-varying; the uniformity over the function class therefore rests on moment and dependence conditions whose verification is not fully detailed for this regime. Because the critical values for the conditional-independence test are obtained directly from this approximation, any gap here is load-bearing for the central claim.

    Authors: We agree that the uniformity of the Gaussian approximation over the function class is load-bearing and that the interaction between the shrinking bandwidth and the time-varying regression requires careful justification in the single-realization setting. Theorem 4.1 is stated under Assumptions 3.1–3.4, which impose moment and dependence conditions that are formulated to be uniform over the relevant function class and to accommodate local estimation with bandwidth h_n → 0. The appendix proof combines maximal inequalities for nonstationary triangular arrays with a blocking argument that explicitly accounts for the effective local sample size of order n h_n. Nevertheless, we acknowledge that sections 3.2–3.3 present these steps at a relatively high level. In the revised manuscript we will add a dedicated paragraph in §3.3 that walks through the verification of uniformity under the stated assumptions, highlighting how the time-varying regression is handled via local smoothing and how the moment conditions scale with the local sample size. This clarification will make the route from assumptions to the critical values fully explicit without changing the theorem statement itself. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper presents a framework for conditional independence testing based on time-varying nonlinear regression, estimation of local long-run covariance matrices of error products, and a distribution-uniform strong Gaussian approximation. These components are described as key technical ingredients drawn from standard statistical tools without any quoted reduction showing that a central result equals its inputs by construction, a fitted parameter renamed as a prediction, or a load-bearing premise justified solely by overlapping self-citation. The abstract and described method indicate an independent construction that does not collapse to tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract, no free parameters, axioms, or invented entities are specified; the approach appears to rely on existing statistical tools without introducing new postulated entities.

pith-pipeline@v0.9.0 · 5669 in / 1168 out tokens · 61592 ms · 2026-05-22T18:00:11.237261+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages · 3 internal anchors

  1. [1]

    The generalization ability of online algorithms for dependent data

    [AD12] Alekh Agarwal and John C. Duchi. “The generalization ability of online algorithms for dependent data”. In:IEEE Transactions on Information Theory 59.1 (2012), pp. 573–587. [ALW13] Pierre Alquier, Xiaoyin Li, and Olivier Wintenberger. “Prediction of time series by statistical learning: general losses and fast rates”. In:Dependence Modeling 1 (2013),...

    work page arXiv 2012

  2. [2]

    Inference for parameters defined by moment inequalities: a recommended moment selection procedure

    [AB12] Donald W.K. Andrews and Panle Jia Barwick. “Inference for parameters defined by moment inequalities: a recommended moment selection procedure”. In: Econometrica: Journal of the Econometric Society80.6 (2012), pp. 2805–

    work page 2012

  3. [3]

    Validity of subsampling and plug-in asymptotic inference for parameters defined by moment inequalities

    [AG09] Donald W.K. Andrews and Patrik Guggenberger. “Validity of subsampling and plug-in asymptotic inference for parameters defined by moment inequalities”. In: Econometric Theory25.3 (2009), pp. 669–709. [AS10] Donald W.K. Andrews and Gustavo Soares. “Inference for parameters defined by moment inequalities using generalized moment selection”. In:Econome...

    work page arXiv 2009

  4. [4]

    Robust inference with knockoffs

    [BCS20] Rina Foygel Barber, Emmanuel J. Candés, and Richard J. Samworth. “Robust inference with knockoffs”. In:The Annals of Statistics48.3 (2020), pp. 1409–

    work page 2020

  5. [5]

    Testing goodness-of-fit and conditional independence with approximate co-sufficient sampling

    [BJ22] Rina Foygel Barber and Lucas Janson. “Testing goodness-of-fit and conditional independence with approximate co-sufficient sampling”. In:The Annals of Statistics 50.5 (2022), pp. 2514–2544. 120 [BM15] Sumanta Basu and George Michailidis. “Regularized estimation in sparse high-dimensional time series models”. In:Annals of Statistics43.4 (2015), pp. 1...

    work page 2022

  6. [6]

    Some new asymptotic theory for least squares series: pointwise and uniform results

    [Bel+15] Alexandre Belloni, Victor Chernozhukov, Denis Chetverikov, and Kengo Kato. “Some new asymptotic theory for least squares series: pointwise and uniform results”. In:Journal of Econometrics186.2 (2015), pp. 345–366. [BH95] Yoav Benjamini and Yosef Hochberg. “Controlling the false discovery rate: a practical and powerful approach to multiple testing...

    work page arXiv 2015

  7. [7]

    Gaussian approxima- tion for non-stationary time series with optimal rate and explicit construction

    121 [BKW24] Soham Bonnerjee, Sayar Karmakar, and Wei Biao Wu. “Gaussian approxima- tion for non-stationary time series with optimal rate and explicit construction”. In: The Annals of Statistics52.5 (2024), pp. 2293–2317. [BRT12] Taoufik Bouezmarni, Jeroen V.K. Rombouts, and Abderrahim Taamouti. “Non- parametric copula-based test for conditional independen...

    work page 2024

  8. [8]

    Panning for gold: model-X knockoffs for high dimensional controlled variable selection

    13140/RG.2.2.36779.31523/1. [Can+18] Emmanuel J. Candés, Yingying Fan, Lucas Janson, and Jinchi Lv. “Panning for gold: model-X knockoffs for high dimensional controlled variable selection”. In: Journal of the Royal Statistical Society Series B: Statistical Methodology 80.3 (2018), pp. 551–577. [CZK24] Abhinav Chakraborty, Jeffrey Zhang, and Eugene Katsevi...

    work page arXiv 2018

  9. [9]

    Cross validation for locally stationary processes

    [DR19] Rainer Dahlhaus and Stefan Richter. “Cross validation for locally stationary processes”. In:Annals of Statistics47.4 (2019), pp. 2145–2173. 123 [DRW19] Rainer Dahlhaus, Stefan Richter, and Wei Biao Wu. “Towards a general theory for nonlinear locally stationary processes”. In:Bernoulli 25.2 (2019), pp. 1013–

    work page 2019

  10. [10]

    Partial association measures and an application to qualitative regression

    [Dau80] J. J. Daudin. “Partial association measures and an application to qualitative regression”. In:Biometrika 67.3 (1980), pp. 581–590. [DN20] Richard A. Davis and Mikkel S. Nielsen. “Modeling of time series using random forests: theoretical developments”. In:Electronic Journal of Statistics(2020), pp. 3644–3671. [DT22] Philip A. Dawid and Ambuj Tewari...

    work page 1980

  11. [11]

    On the partial autocorrelation function for locally stationary time series: characterization, estimation and inference

    [DZ25] Xiucai Ding and Zhou Zhou. “On the partial autocorrelation function for locally stationary time series: characterization, estimation and inference”. In: Biometrika 112.2 (2025). [DZ21] Xiucai Ding and Zhou Zhou. “Simultaneous sieve inference for time- inhomogeneousnonlineartimeseriesregression”.arXivpreprintarXiv:2112.08545

    work page arXiv 2025

  12. [12]

    A permutation-based kernel conditional independence test

    [Dor+14] Gary Doran, Krikamol Muandet, Kun Zhang, and Bernhard Schölkopf. “A permutation-based kernel conditional independence test”. In:Proceedings of the 124 Thirtieth Conference on Uncertainty in Artificial Intelligence(2014), pp. 132–

    work page 2014

  13. [13]

    Graphical modelling of multivariate time series

    [Eic12] Michael Eichler. “Graphical modelling of multivariate time series”. In:Proba- bility Theory and Related Fields153 (2012), pp. 233–268. [FFX20] Jianqing Fan, Yang Feng, and Lucy Xia. “A projection-based conditional dependencemeasurewithapplicationstohigh-dimensionalundirectedgraphical models”. In:Journal of Econometrics218.1 (2020), pp. 119–139. [F...

    work page 2012

  14. [14]

    Cyclostationarity: half a century of research

    [GNP06] William A. Gardner, Antonio Napolitano, and Luigi Paura. “Cyclostationarity: half a century of research”. In:Signal Processing86.4 (2006), pp. 639–697. [Goe20] Benjamin Goehry. “Random forests for time-dependent processes”. In:ESAIM: Probability and Statistics24 (2020), pp. 801–826. [GMR93] Christian Gourieroux, Alain Monfort, and Eric Renault. “I...

    work page 2006

  15. [15]

    Confidence intervals for partially identified parameters

    [IM04] Guido W. Imbens and Charles F. Manski. “Confidence intervals for partially identified parameters”. In:Econometrica: Journal of the Econometric Society 72.6 (2004), pp. 1845–1857. [KKR24] Iden Kalemaj, Shiva Kasiviswanathan, and Aaditya Ramdas. “Differentially private conditional independence testing”. In:International Conference on Artificial Intel...

    work page 2004

  16. [16]

    Rates of uniform convergence of empirical means with mixing processes

    [KV02] Rajeeva L. Karandikar and Mathukumalli Vidyasagar. “Rates of uniform convergence of empirical means with mixing processes”. In:Statistics and Probability Letters58.3 (2002), pp. 297–307. [KW20] Sayar Karmakar and Wei Biao Wu. “Optimal Gaussian approximation for multiple time series”. In: 30.3 (2020), pp. 1399–1417. [Kas18] Maximilian Kasy. “Uniform...

    work page 2002

  17. [17]

    Post-selection inference

    [KKK22] Arun K. Kuchibhotla, John E. Kolassa, and Todd A. Kuffner. “Post-selection inference”. In:Annual Review of Statistics and Its Application9.1 (2022), pp. 505–527. [KBW23] Arun Kumar Kuchibhotla, Sivaraman Balakrishnan, and Larry Wasserman. “Median regularity and honest inference”. In:Biometrika 110.3 (2023), pp. 831–

    work page 2022

  18. [18]

    Generalization bounds for non- stationary mixing processes

    [KM17] Vitaly Kuznetsov and Mehryar Mohri. “Generalization bounds for non- stationary mixing processes”. In:Machine Learning 106.1 (2017), pp. 93–

    work page 2017

  19. [19]

    Generalization bounds for time series prediction with non-stationary processes

    [KM14] Vitaly Kuznetsov and Mehryar Mohri. “Generalization bounds for time series prediction with non-stationary processes”. In:Algorithmic Learning Theory25 (2014), pp. 260–274. [KM15] Vitaly Kuznetsov and Mehryar Mohri. “Learning theory and algorithms for forecasting non-stationary time series”. In:Advances in Neural Information Processing Systems28 (20...

    work page 2014

  20. [20]

    Honest confidence regions for nonparametric regression

    [Li89] Ker-Chau Li. “Honest confidence regions for nonparametric regression”. In: The Annals of Statistics17.3 (1989), pp. 1001–1008. [Li+11] Lingling Li, Eric Tchetgen Tchetgen, Aad van der Vaart, and James M. Robins. “Higher order inference on a treatment effect under low regularity conditions”. In: Statistics and Probability Letters81.7 (2011), pp. 821...

    work page arXiv 1989

  21. [21]

    Learning the structure of a nonstationary vector autoregression

    129 [MS19] Daniel Malinsky and Peter Spirtes. “Learning the structure of a nonstationary vector autoregression”. In:International Conference on Artificial Intelligence and Statistics89 (2019), pp. 2986–2994. [Man+24] Georg Manten, Cecilia Casolo, Emilio Ferrucci, Søren Wengel Mogensen, Cristo- pher Salvi, and Niki Kilbertus. “Signature kernel conditional ...

    work page arXiv 2019

  22. [22]

    Distribution-free learning of Bayesian network structure in continuous domains

    [Mar05] Dimitris Margaritis. “Distribution-free learning of Bayesian network structure in continuous domains”. In:AAAI 5 (2005), pp. 825–830. [McF89] Daniel McFadden. “A method of simulated moments for estimation of discrete response models without numerical integration”. In:Econometrica: Journal of the Econometric Society57.5 (1989), pp. 995–1026. [Mie24...

    work page arXiv 2005

  23. [23]

    Cross-Fitting and Fast Remainder Rates for Semiparametric Estimation

    [MS23] Fabian Mies and Ansgar Steland. “Sequential Gaussian approximation for nonstationary time series in high dimensions”. In:Bernoulli 29.4 (2023), pp. 3114–3140. [MK20] Mehryar Mohri and Vitaly Kuznetsov. “Discrepancy-based theory and algo- rithms for forecasting non-stationary time series”. In:Annals of Mathematics and Artificial Intelligence88.4 (20...

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  24. [24]

    Minimax optimal conditional independence testing

    [NBW21] Matey Neykov, Sivaraman Balakrishnan, and Larry Wasserman. “Minimax optimal conditional independence testing”. In:The Annals of Statistics49.4 (2021), pp. 2151–2177. [Niu+24] Ziang Niu, Abhinav Chakraborty, Oliver Dukes, and Eugene Katsevich. “Rec- onciling model-X and doubly robust approaches to conditional independence testing”. In:The Annals of...

    work page 2021

  25. [25]

    Lasso regression in sparse linear model with phi-mixing errors

    [PZZ23] Ling Peng, Yan Zhu, and Wenxuan Zhong. “Lasso regression in sparse linear model with phi-mixing errors”. In:Metrika 86.1 (2023), pp. 1–26. [Pet+14] Jonas Peters, Joris M. Mooij, Dominik Janzing, and Bernhard Schölkopf. “Causal discovery with continuous additive noise models”. In:Journal of Machine Learning Research15.58 (2014), pp. 2009–2053. [Pis...

    work page 2023

  26. [26]

    A Scalable Conditional Independence Test for Nonlinear, Non-Gaussian Data

    [QKC15] Christopher J. Quinn, Negar Kiyavash, and Todd P. Coleman. “Directed information graphs”. In:IEEE Transactions on Information Theory61.12 (2015), pp. 6887–6909. [Ram14] Joseph D. Ramsey. “A scalable conditional independence test for nonlinear, non-gaussian data”. arXiv preprint arXiv:1401.5031

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  27. [27]

    Central limit theorems for local martingales

    131 [Reb80] Rolando Rebolledo. “Central limit theorems for local martingales”. In: Zeitschrift für Wahrscheinlichkeitstheorie und verwandte Gebiete51.3 (1980), pp. 269–286. [RGR22] Nicolas-Domenic Reiter, Andreas Gerhardus, and Jakob Runge. “Causal infer- ence for temporal patterns”. arXiv preprint arXiv:2205.15149

    work page arXiv 1980

  28. [28]

    Improving the bandwidth-free inference methods by prewhitening

    [RS13] Yeonwoo Rho and Xiaofeng Shao. “Improving the bandwidth-free inference methods by prewhitening”. In:Journal of Statistical Planning and Inference 143.11 (2013), pp. 1912–1922. [RWG19] Alessandro Rinaldo, Larry Wasserman, and Max G’Sell. “Bootstrapping and sample splitting for high-dimensional, assumption-lean inference”. In:The Annals of Statistics...

    work page 2013

  29. [29]

    Causal network reconstruction from time series: from theoretical assumptions to practical estimation

    [Run18a] Jakob Runge. “Causal network reconstruction from time series: from theoretical assumptions to practical estimation”. In:Chaos: An Interdisciplinary Journal of Nonlinear Science28.7 (2018). [Run18b] Jakob Runge. “Conditional independence testing based on a nearest-neighbor estimator of conditional mutual information”. In:International Conference o...

    work page 2018

  30. [30]

    Nonparametric tests of conditional indepen- dence for time series

    134 [SW21] Xiaojun Song and Haoyu Wei. “Nonparametric tests of conditional indepen- dence for time series”. arXiv preprint arXiv:2110.04847

    work page arXiv

  31. [31]

    Consistency of support vector machines for forecasting the evolution of an unknown ergodic dynamical system from observations with unknown noise

    [SA09] Ingo Steinwart and Marian Anghel. “Consistency of support vector machines for forecasting the evolution of an unknown ergodic dynamical system from observations with unknown noise”. In:Annals of Statistics37.2 (2009), pp. 841–

    work page 2009

  32. [32]

    Learning from dependent obser- vations

    [SHS09] Ingo Steinwart, Don Hush, and Clint Scovel. “Learning from dependent obser- vations”. In:Journal of Multivariate Analysis100.1 (2009), pp. 175–194. [SW07] Liangjun Su and Halbert White. “A consistent characteristic function-based test for conditional independence”. In:Journal of Econometrics141.2 (2007), pp. 807–834. [SW08] Liangjun Su and Halbert...

    work page arXiv 2009

  33. [33]

    Lasso guarantees for beta- mixing heavy-tailed time series

    [WLT20] Kam Chung Wong, Zifan Li, and Ambuj Tewari. “Lasso guarantees for beta- mixing heavy-tailed time series”. In:The Annals of Statistics48.2 (2020), pp. 1124–1142. [Wu11] Wei Biao Wu. “Asymptotic theory for stationary processes”. In:Statistics and its Interface4.2 (2011), pp. 207–226. [Wu05] Wei Biao Wu. “Nonlinear system theory: another look at depe...

    work page 2020

  34. [34]

    Bootstrapping High Dimensional Time Series

    Elsevier, 2012, pp. 187–209. [WW16] Wei-Biao Wu and Ying Nian Wu. “Performance bounds for parameter estimates of high-dimensional linear models with correlated errors”. In:Electronic Journal of Statistics10.1 (2016), pp. 352–379. [WZ24] Weichi Wu and Zhou Zhou. “Multiscale jump testing and estimation under complex temporal dynamics”. In:Bernoulli 30.3 (20...

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  35. [35]

    Another look at bandwidth-free inference: a sample splitting approach

    [ZS24] Yi Zhang and Xiaofeng Shao. “Another look at bandwidth-free inference: a sample splitting approach”. In:Journal of the Royal Statistical Society Series B: Statistical Methodology86.1 (2024), pp. 246–272. [ZZZ22] Yeqing Zhou, Yaowu Zhang, and Liping Zhu. “A Projective Approach to Conditional Independence Test for Dependent Processes”. In:Journal of ...

    work page 2024