Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Partial Decomposition of Granger Causality splits multivariate spectral causality into unique, redundant and synergistic atoms that distinguish syncope patients from healthy controls under postural stress.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-15 13:07 UTC pith:6GSXYKKI

load-bearing objection Solid spectral PID of multivariate GC that organizes known pairwise/conditional intuitions and yields clean tilt contrasts; the min-redundancy choice is a modeling decision, not a theorem, and the promised simulations are missing from the body. the 3 major comments →

arxiv 2603.07634 v2 pith:6GSXYKKI submitted 2026-03-08 stat.ME physics.data-an

Dissecting Spectral Granger Causality through Partial Information Decomposition

Luca Faes , Gorana Mijatovic , Riccardo Pernice , Daniele Marinazzo , Sebastiano Stramaglia , Yuri Antonacci This is my paper
classification stat.ME physics.data-an
keywords Granger causalitypartial information decompositionspectral analysishigh-order interactionsstate-space modelsnetwork physiologycardiovascular controlcerebrovascular autoregulation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper introduces Partial Decomposition of Granger Causality (PDGC), a method that takes the familiar multivariate Granger causality from several driver time series to a target and splits it into unique, redundant and synergistic pieces. It does so by embedding frequency-domain Granger measures inside the partial-information-decomposition lattice, defining redundancy at each frequency as the minimum of the drivers’ spectral causalities and recovering the atoms by Möbius inversion. Because the construction lives in the spectral domain, the same atoms can be integrated over any physiologically meaningful band or over the whole spectrum. On cardiovascular and cerebrovascular series the resulting unique, redundant and synergistic components show opposite responses to head-up tilt in healthy subjects versus patients prone to neurally-mediated syncope, revealing high-order control patterns that ordinary pairwise or full multivariate Granger causality leave opaque. The method therefore supplies a practical, frequency-resolved language for high-order directed interactions among oscillatory processes.

Core claim

Multivariate spectral Granger causality from a set of drivers to a target can be decomposed, via a pointwise min-redundancy function and Möbius inversion on the PID lattice, into non-negative unique, redundant and synergistic spectral atoms that integrate exactly to the classical time-domain Granger measures; these atoms, when evaluated on physiological series, expose distinctive high-order causal reorganizations under postural stress that separate syncope patients from matched controls.

What carries the argument

The spectral redundant Granger causality f^∩_{X_α→Y}(ω) := min_j f_{X_αj→Y}(ω), together with its full-band integral and the subsequent Möbius inversion that yields the atomic GCs; the construction is realized by state-space models of the VAR process so that reduced models remain exact.

Load-bearing premise

Redundancy is defined as the pointwise minimum of the spectral Granger measures of the atom’s source subsets, and all measures are obtained from linear state-space models; if that minimum misrepresents shared causal information, or if important nonlinear causal structure is present, the unique/redundant/synergistic split can fail even when full multivariate GC is large.

What would settle it

On a controlled linear Gaussian network whose ground-truth unique, redundant and synergistic causal strengths are known, compute PDGC; if the recovered atoms deviate systematically from those known strengths, or if the same atoms change sign or disappear under mild non-Gaussian driving noise while full GC remains unchanged, the decomposition is falsified.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Any existing multivariate spectral GC analysis can be re-run with PDGC to report unique, redundant and synergistic contributions band by band.
  • In network physiology the unique SAP→HP and the redundant/synergistic MAP/HP→MCBV components become candidate biomarkers of baroreflex and cerebral-autoregulation dysfunction under orthostatic stress.
  • Because the atoms integrate exactly across frequency, the same decomposition can be restricted to any a-priori band of interest without recomputing the underlying models.
  • The method supplies a directed, frequency-resolved counterpart to existing undirected high-order information measures, enabling comparison of pairwise versus multi-body causal routes in oscillatory networks.

Where Pith is reading between the lines

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

  • The same spectral-min construction could be applied to other directed spectral measures (e.g., directed transfer function or partial directed coherence) to obtain analogous unique/redundant/synergistic decompositions.
  • If the linear-Gaussian assumption is relaxed by replacing the state-space GC with a model-free spectral transfer entropy, the resulting atoms would test whether the physiological findings survive nonlinear interactions.
  • The opposite tilt responses of CV versus CB high-order atoms suggest that a joint PDGC of the full five-variable network might reveal compensatory loops that are invisible when the two sub-networks are analysed separately.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The manuscript introduces Partial Decomposition of Granger Causality (PDGC), which embeds multivariate spectral Granger causality (GC) computed from state-space models into the partial information decomposition (PID) framework. Multivariate GC from a set of drivers X to a target Y is dissected into unique, redundant, and synergistic atoms by defining a spectral redundant GC as the pointwise minimum of the spectral GCs of the atom’s source subsets (Eq. 13), recovering atoms by Möbius inversion (Eqs. 14–15), and coarse-graining into unique/redundant/synergistic components (Eqs. 16–17). Whole-band and band-limited (LF/HF) integrals recover time-domain measures. The method is applied to cardiovascular (RESP, SAP → HP) and cerebrovascular (RESP, MAP, HP → MCBV) networks in syncope patients versus controls at rest and during head-up tilt, reporting blunted unique/redundant SAP o HP responses in patients and elevated LF redundant/synergistic MAP/HP o MCBV effects that are interpreted as markers of autonomic and autoregulatory dysfunction.

Significance. If the construction is sound, PDGC supplies a directed, frequency-resolved high-order causality tool that is more interpretable than pairwise or conditional GC alone and is computationally convenient via closed SS reduced models. The physiological application is of genuine interest for network physiology: it refines known baroreflex and cerebral-autoregulation findings by attributing tilt responses to unique versus higher-order components in specific bands. Strengths include the SS formulation that avoids infinite-order VAR truncation, the spectral-integration property that links time- and frequency-domain PIDs, and the use of surrogate testing plus non-parametric group comparisons. The abstract’s claim of benchmark validation, if present and rigorous, would further strengthen the contribution for data-driven network science.

major comments (3)
  1. [Abstract / §I] Abstract and §I promise “validation on benchmark simulations” showing that unique/redundant/synergistic GC “reflect the underlying causal mechanisms and are computationally reliable.” No such simulations appear in the manuscript (Methods, Results, or appendices). Without ground-truth recovery experiments (e.g., known redundant/synergistic VAR or SS networks), the central claim that the atoms correctly dissect causal mechanisms remains untested.
  2. [§IIB, Eq. (13)] Eq. (13) defines spectral redundant GC as the pointwise minimum of the spectral GCs of the atom’s source subsets. This is a modeling choice (I_min-style), not a theorem. Different redundancy functions can re-partition the same full GC into different unique/redundant/synergistic values. The manuscript never demonstrates that the reported physiological conclusions (tilt-induced unique SAP o HP rise in controls; LF redundant/synergistic MAP/HP o MCBV rise in SYNC) are invariant under an alternative redundancy definition, nor does it compare against other PID candidates. Because all coarse-graining (Fig. 1, Eq. 16) and band-limited interpretations inherit this definition, sensitivity analysis is load-bearing.
  3. [§IIC / §IIIB] All spectral GC atoms are obtained from linear Gaussian VAR/SS models of short (N=250) physiological series. The manuscript does not assess robustness to mild nonlinearity, non-Gaussian innovations, or model-order misspecification, all of which are common in cardiovascular/cerebrovascular data. If the linear assumption fails, the unique/redundant/synergistic split can be distorted even when full multivariate GC is nonzero. At minimum, a brief residual diagnostic or comparison with a nonlinear GC baseline on the same data would be needed to support the physiological claims in §IIID.
minor comments (5)
  1. [§IIIC] Figure numbering is inconsistent: the CV results are introduced as “Fig. 4” but the caption and subsequent text refer to Fig. 3; CB results are also labeled Fig. 4. Correct the labels and cross-references.
  2. [Abstract] The abstract states that validation “demonstrates … computational reliability,” yet no numerical stability, bias, or variance results are reported. Either add a short computational-reliability subsection or soften the abstract claim.
  3. [§IIA–IIC] Notation for the restricted process Z and the Cholesky-adjusted transfer function H̃ versus H is dense; a short table of symbols or an expanded sentence after Eq. (20) would help readers implement the method.
  4. [§IIIB] Surrogate significance is assessed at the single-subject level with IAAFT (100 surrogates, 95th percentile). Clarify whether multiple-comparison correction across atoms/bands was considered, or state that none was applied.
  5. [§IV] Several self-citations to the authors’ PIRD and HOI papers are appropriate for the building blocks, but a brief explicit comparison of PDGC versus SURD and PIRD (beyond the one-sentence claim in §IV) would better position the novelty.

Circularity Check

1 steps flagged

Minor self-citation of authors' own PIRD redundancy ansatz supplies the min definition; PDGC atoms and physiological contrasts are not forced by construction from that choice alone.

specific steps
  1. ansatz smuggled in via citation [Sect. II.B, Eqs. 12–13 and surrounding text]
    "Here, we follow the rationale of the recently proposed partial information rate decomposition [19], employing frequency-domain expansion and defining the redundant GC as the full-frequency integral of the so-called spectral redundant GC: F∩Xα→Y:=1/π∫π0f∩Xα→Y(ω)dω. The pointwise spectral redundant GC is in turn defined, at each normalized angular frequency ω∈[0,π], by taking the minimum of the spectral GCs directed to the target process from each subset of drivers composing the analyzed atom: f∩Xα→Y(ω):=minj=1,...,JfXαj→Y(ω)"

    The load-bearing definition that partitions full spectral GC into unique/redundant/synergistic atoms is imported wholesale from the authors' own prior PIRD paper [19] rather than derived or compared against alternatives inside the present manuscript. All subsequent Möbius atoms, coarse-graining (Eq. 16) and band-limited physiological interpretations inherit this min ansatz by construction; a different redundancy function would re-label the same full GC without changing the underlying linear spectral estimates.

full rationale

The paper is a methodological proposal that defines PDGC by embedding spectral GC (via SS/VAR models) into the PID lattice, with the sole non-standard modeling choice being the pointwise-min redundancy function taken from the authors' concurrent PIRD work. That choice is an explicit ansatz, not a uniqueness theorem, and the subsequent Möbius inversion and coarse-graining are pure algebra that hold by construction for any redundancy function. The empirical claims (group differences under tilt) are ordinary statistical comparisons of the resulting measures on real data; they do not reduce to a fitted free parameter or to a self-citation chain that already contains the syncope result. No equation equates a 'prediction' to its own input. The self-citation burden is therefore present but not load-bearing for any claimed derivation, yielding only a score of 2.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 2 invented entities

The central claim rests on standard GC/PID/SS mathematics, a domain choice of linear VAR models for beat-to-beat physiological series, an ad-hoc-to-paper redundancy definition (pointwise min of spectral GCs), and a few analysis hyperparameters (model order via BIC, LF/HF cutoffs, surrogate threshold). PDGC and its spectral atoms are the main invented constructs; they inherit falsifiability only through simulation and empirical behavior, not through an external conserved quantity or predicted physical constant.

free parameters (4)
  • VAR model order p
    Chosen in range 3–12 by BIC for each multivariate series; affects all spectral GC and PDGC estimates.
  • LF and HF integration bands
    Fixed physiological cutoffs LF [0.03,0.15] Hz and HF [0.15,0.4] Hz used for band-limited PDGC; band edges are conventional but analysis-defining.
  • Surrogate significance threshold
    IAAFT percentile test at 5% (95th percentile of 100 surrogates) decides single-subject significance of each GC component.
  • Trend-removal cutoff
    Zero-phase AR low-pass at 0.0156 cycles/beat before analysis; preprocessing choice that can alter low-frequency GC content.
axioms (5)
  • standard math Time- and frequency-domain Granger causality as predictive variance reduction under linear VAR/SS models (Geweke-style spectral GC with instantaneous-effect handling).
    Used throughout §IIA–IIC as the base measure being decomposed.
  • standard math Partial information decomposition lattice with redundancy function and Möbius inversion recovering unique/redundant/synergistic atoms.
    Williams–Beer PID structure and consistency equations invoked in §IIB.
  • ad hoc to paper Spectral redundant GC is the pointwise minimum of the spectral GCs of the atom’s source subsets, and whole-band integration recovers a valid time-domain PID.
    Eqs. 12–14; follows the authors’ recent PIRD rationale but is a specific redundancy choice, not forced by PID axioms alone.
  • domain assumption Beat-to-beat cardiovascular and cerebrovascular variability is adequately described by finite-order linear Gaussian state-space dynamics for causal inference under rest and tilt.
    Implicit in VAR/SS identification and spectral GC for HP, SAP/MAP, RESP, MCBV in §III.
  • domain assumption IAAFT surrogates that preserve spectra but destroy cross-spectra provide a valid null for significance of directed GC/PDGC measures.
    Used for single-subject significance in §IIIB.
invented entities (2)
  • Partial Decomposition of Granger Causality (PDGC) no independent evidence
    purpose: Name and framework for dissecting multivariate GC into unique, redundant, and synergistic causal atoms in time and frequency domains.
    Core proposed tool; not an external physical entity, but a new named measure family whose validity is internal to the chosen redundancy and linear models.
  • Spectral redundant GC and spectral atomic GCs no independent evidence
    purpose: Frequency-local redundancy and PID atoms of Granger causality used to form band-limited unique/redundant/synergistic measures.
    Defined by Eqs. 13–15; independent evidence would require external benchmarks or open simulation suites, which are claimed but not shown in the body.

pith-pipeline@v1.1.0-grok45 · 20437 in / 3744 out tokens · 37492 ms · 2026-07-15T13:07:11.593780+00:00 · methodology

0 comments
read the original abstract

Granger causality (GC), a popular statistical method for the inference of directional influences between time series measured from a complex network, is sensitive to high-order (non-pairwise) interactions which fundamentally shape the collective network dynamics. This work introduces Partial Decomposition of Granger Causality (PDGC), a tool eliciting redundant and synergistic causal interactions in the pattern of information flow between the subsystems of physiological networks. The tool exploits the framework of partial information decomposition to dissect the multivariate GC from a set of driver random processes to a target process into unique effects carried exclusively by each driver, redundant effects carried identically by more drivers, and synergistic effects carried jointly by some drivers but not by any of them individually. Computation is based on multivariate state-space models expanded in the frequency domain to assess PDGC both in specific bands of physiological interest and in the time domain after whole-band integration. The validation on benchmark simulations demonstrates that the measures of unique, redundant, and synergistic GC reflect the underlying causal mechanisms and are computationally reliable. The application to arterial pressure, respiration, cerebral blood velocity and heart period variability reveals striking differences in the response to postural stress of patients prone to neurally-mediated syncope compared to healthy controls. The extraction of high-order causality patterns from the spectral GC favors dissecting the mechanisms of causal influence underlying multivariate interactions among oscillatory processes in many data-driven applications of network science.

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Partial Effective Information Decomposition for Synergistic Causality

    stat.ML 2026-05 unverdicted novelty 6.0

    PEID decomposes the causal effect of multiple sources on a target under maximum-entropy interventions into unique and synergistic information, enabling hyperedge causal graphs and downward causation analysis.

Reference graph

Works this paper leans on

46 extracted references · 1 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Time-domain formulation Let us consider a discrete-time, real-valuedM- dimensional random processS={S n}n∈Z,S n ∈R M, which is partitioned into a scalar target process and N=M−1driver processes:S n ={X n, Yn},X n ∈ RN , Yn ∈R. In this frame, the best least-squares pre- diction of the present state of the target process,Yn, given the entire past informatio...

  2. [2]

    The two innovations are generally correlated (Σ12 = Σ⊺ 21 = E[W1,nW2,n]̸= 0) due to the instantaneous interactions between the drivers and the target

    Frequency-domain formulation To compute GC in the frequency domain, we adapt the approach proposed in [4] to our setting, first parti- tioning the VAR sub-model (4) as: Xαj ,n = ∞X k=1 B11,kXαj ,n−k + ∞X k=1 B12,kYn−k +W 1,n (5a) Yn = ∞X k=1 B21,kXαj ,n−k + ∞X k=1 b12,kYn−k +W 2,n, (5b) to evidence the VAR representations of the driver and target processe...

  3. [3]

    PDGC Analysis of Cardiovascular Interactions Fig. 4 reports the results of the PDGC analysis of CV interactions, performed by decomposing the Granger- causal effects of RESP and SAP on HP in both non- SYNC and SYNC groups and considering both whole- band and band-specific measures. The overall predic- tive information transferred from RESP and SAP to HP (...

  4. [4]

    PDGC Analysis of Cerebrovascular Interactions Fig. 4 reports the results of the PDGC analysis of CB interactions, performed by decomposing the Granger- causal effects of RESP, MAP and HP on MCBV in both nonSYNC and SYNC groups and considering both whole-band and band-specific measures. The full GC (Fig. 4a,b, left panels) was high and statistically signif...

  5. [5]

    Lehnertz, T

    K. Lehnertz, T. Bröhl, and T. Rings, Frontiers in Phys- iology11, 598694 (2020)

    2020

  6. [6]

    Porta and L

    A. Porta and L. Faes, Proceedings of the IEEE104, 282 (2015)

    2015

  7. [7]

    C. W. Granger, Econometrica: journal of the Econo- metric Society , 424 (1969)

    1969

  8. [8]

    Geweke, Journal of the American statistical associa- tion77, 304 (1982)

    J. Geweke, Journal of the American statistical associa- tion77, 304 (1982)

    1982

  9. [9]

    Barnett, A

    L. Barnett, A. B. Barrett, and A. K. Seth, Physical review letters103, 238701 (2009)

    2009

  10. [10]

    L.A.BaccaláandK.Sameshima,Biologicalcybernetics 84, 463 (2001)

    2001

  11. [11]

    Porta, P

    A. Porta, P. Castiglioni, M. Di Rienzo, T. Bassani, V. Bari, L. Faes, G. Nollo, A. Cividjan, and L. Quintin, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences371, 20120161 (2013)

    2013

  12. [12]

    Stramaglia, L

    S. Stramaglia, L. Angelini, G. Wu, J. M. Cortes, L. Faes, and D. Marinazzo, IEEE Transactions on Biomedical Engineering63, 2518 (2016)

    2016

  13. [13]

    Porta, F

    A. Porta, F. Gelpi, V. Bari, B. Cairo, B. De Maria, D. Tonon, G. Rossato, and L. Faes, Medical & Biolog- ical Engineering & Computing61, 3141 (2023)

    2023

  14. [14]

    Sparacino, Y

    L. Sparacino, Y. Antonacci, C. Barà, A. Valenti, A. Porta, and L. Faes, IEEE Transactions on Biomed- ical Engineering71, 1454 (2023)

    2023

  15. [15]

    Pernice, L

    R. Pernice, L. Sparacino, V. Bari, F. Gelpi, B. Cairo, G. Mijatovic, Y. Antonacci, D. Tonon, G. Rossato, M. Javorka,et al., Autonomic Neuroscience242, 103021 (2022)

    2022

  16. [16]

    F. E. Rosas, P. A. Mediano, H. J. Jensen, A. K. Seth, A. B. Barrett, R. L. Carhart-Harris, and D. Bor, PLoS computational biology16, e1008289 (2020)

    2020

  17. [17]

    F. E. Rosas, P. A. Mediano, A. I. Luppi, T. F. Varley, J. T. Lizier, S. Stramaglia, H. J. Jensen, and D. Mari- nazzo, Nature Physics18, 476 (2022)

    2022

  18. [18]

    L. Faes, G. Mijatovic, L. Sparacino, and A. Porta, IEEE Transactions on Biomedical Engineering72, 3515 (2025)

    2025

  19. [19]

    P. L. Williams and R. D. Beer, arXiv preprint arXiv:1004.2515 (2010)

    Pith/arXiv arXiv 2010

  20. [20]

    F. E. Rosas, P. A. Mediano, M. Gastpar, and H. J. Jensen, Physical Review E100, 032305 (2019)

    2019

  21. [21]

    T. F. Varley, M. Pope, J. Faskowitz, and O. Sporns, Communications biology6, 451 (2023)

    2023

  22. [22]

    L. Faes, G. Mijatovic, Y. Antonacci, R. Pernice, C. Barà, L. Sparacino, M. Sammartino, A. Porta, D. Marinazzo, and S. Stramaglia, IEEE Transactions on Signal Processing70, 5766 (2022)

    2022

  23. [23]

    L. Faes, L. Sparacino, G. Mijatovic, Y. Antonacci, L. Ricci, D. Marinazzo, and S. Stramaglia, Physical Review Letters135, 187401 (2025)

    2025

  24. [24]

    Scagliarini, L

    T. Scagliarini, L. Sparacino, L. Faes, D. Marinazzo, and S. Stramaglia, Frontiers in Network Physiology3, 1335808 (2024)

    2024

  25. [25]

    J. F. Geweke, Journal of the American Statistical As- sociation79, 907 (1984)

    1984

  26. [26]

    A. B. Barrett, L. Barnett, and A. K. Seth, Physical Re- view E—Statistical, Nonlinear, and Soft Matter Physics 81, 041907 (2010)

    2010

  27. [27]

    Stramaglia, J

    S. Stramaglia, J. M Cortes, and D. Marinazzo, New Journal of Physics16, 105003 (2014)

    2014

  28. [28]

    Barnett and A

    L. Barnett and A. K. Seth, Physical Review E91, 040101 (2015)

    2015

  29. [29]

    Gelpi, V

    F. Gelpi, V. Bari, B. Cairo, B. De Maria, D. Tonon, G. Rossato, L. Faes, and A. Porta, Autonomic Neuro- 12 science237, 102920 (2022)

    2022

  30. [30]

    L. Faes, S. Erla, A. Porta, and G. Nollo, Philosophi- cal Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences371(2013)

    2013

  31. [31]

    A. J. Gutknecht, M. Wibral, and A. Makkeh, Proceed- ings of the Royal Society A: Mathematical, Physical and Engineering Sciences477(2021)

    2021

  32. [32]

    Jansma, P

    A. Jansma, P. A. Mediano, and F. E. Rosas, Physical Review Research7, 033049 (2025)

    2025

  33. [33]

    Sparacino, H

    L. Sparacino, H. Pinto, C. Barà, Y. Antonacci, R. Per- nice, A. P. Rocha, and L. Faes, Frontiers in Network Physiology5, 1687132 (2025)

    2025

  34. [34]

    L. Faes, S. Stramaglia, and D. Marinazzo, F1000Research6, 1710 (2017)

    2017

  35. [35]

    T. J. Mullen, M. L. Appel, R. Mukkamala, J. M. Math- ias, and R. J. Cohen, American Journal of Physiology- Heart and Circulatory Physiology272, H448 (1997)

    1997

  36. [36]

    A. J. Ocon, J. Kulesa, D. Clarke, I. Taneja, M. S. Medow, and J. M. Stewart, American Journal of Physiology-Heart and Circulatory Physiology297, H2084 (2009)

    2009

  37. [37]

    M. S. Medow, A. T. Del Pozzi, Z. R. Messer, C. Ter- illi, and J. M. Stewart, Frontiers in physiology5, 234 (2014)

    2014

  38. [38]

    L. Faes, S. Erla, and G. Nollo, Computational and mathematical methods in medicine2012, 140513 (2012)

    2012

  39. [39]

    Schreiber and A

    T. Schreiber and A. Schmitz, Physical review letters77, 635 (1996)

    1996

  40. [40]

    M. A. Cohen and J. A. Taylor, The Journal of physiol- ogy542, 669 (2002)

    2002

  41. [41]

    L. Faes, G. Nollo, and A. Porta, Journal of applied physiology114, 1657 (2013)

    2013

  42. [42]

    L. Faes, D. Marinazzo, A. Montalto, and G. Nollo, IEEE Transactions on Biomedical Engineering61, 2556 (2014)

    2014

  43. [43]

    L. Faes, A. Porta, G. Nollo, and M. Javorka, Entropy 19, 5 (2016)

    2016

  44. [44]

    L. Faes, L. Widesott, M. Del Greco, R. Antolini, and G. Nollo, IEEE transactions on biomedical engineering 53, 65 (2005)

    2005

  45. [45]

    J. M. Karemaker, Journal of applied physiology106, 1742 (2009)

    2009

  46. [46]

    Martínez-Sánchez, G

    Á. Martínez-Sánchez, G. Arranz, and A. Lozano- Durán, Nature Communications15, 9296 (2024)

    2024