Pith. sign in

REVIEW 3 major objections 3 minor 64 references

Low-Rank Structured Nonparametric Prediction of Instantaneous Volatility

T0 review · 3 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A nonparametric low-rank completion identity predicts the remaining intraday volatility path without specifying any volatility model.

desk verdict Genuinely new low-rank completion idea for same-day volatility prediction, but the theory's martingale condition does not cover the two-sided kernel used in the numerics. read the letter →

arxiv 2507.22173 v1 pith:MX43CLZE submitted 2025-07-29 econ.EM

Sung Hoon Choi , Donggyu Kim This is my paper · ORCID
classification econ.EM MSC 62G0562M1062H12
keywords instantaneousvolatilityhigh-frequencydatalow-rankmatrixcompletionnonparametricpredictionintradayValue-at-Risksingularvaluedecomposition
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Volatility forecasters usually must commit to a model: an autoregression, a GARCH-type recursion, or a machine-learning rule. This paper claims that the remaining intraday volatility path of the current trading day can be predicted without any such parametric commitment, using only the assumption that the day-by-intraday instantaneous volatility matrix is a low-rank signal plus noise. The remaining part of today's row of that matrix is completed from the observed part of today's row and the historical rows through a fixed algebraic identity. The paper proves the resulting SIP estimator is consistent with an explicit element-wise rate, and reports that it beats AR, SARIMA, HAR-D, XGBoost, and factor or semiparametric benchmarks in out-of-sample prediction and Value-at-Risk tests on 12 ETFs. If the claim holds, practitioners get a misspecification-robust forecasting tool that works with a small early slice of the day.

What carries the argument

The load-bearing object is the rank-$r$ block-completion identity $A_{22}=A_{21}(A_{11})^{\dagger}A_{12}$, which expresses the unobserved future volatility block through the observed blocks $A_{11}$, $A_{21}$, and $A_{12}$. The SIP estimator is the plug-in version $\tilde\Sigma_{22}=\hat\Sigma_{21}\hat V_{11}(\hat U_{11}^{\top}\hat\Sigma_{11}\hat V_{11})^{-1}\hat U_{11}^{\top}\hat\Sigma_{12}$, built from pre-averaged spot volatility estimates and from singular vectors of the observed row and column blocks. The proof machinery is entrywise singular-vector perturbation theory, which controls how much estimated singular vectors deviate from the true ones entry by entry, together with sub-Gaussian concentration; this splits the prediction error into the spot-estimation component, the cost of learning interday dynamics, and the cost of learning intraday patterns.

What would settle it

Use the implemented estimator (4.1) on simulated data with a known volatility path, regress the errors $\hat c_{i,j}-c_{i,j}$ on information available before time $t_j$, and test whether the conditional mean is zero; a nonzero mean at a detectable order would falsify the martingale-difference assumption for that estimator and leave Theorem 3.1's rate unsupported.

Watch

Extended reading notes

Core claim

Under the model $\Sigma_{D,n}=U\Lambda V^{\top}+E$, the conditional expectation of the remaining volatility vector on day $D$ is the row block $A_{22}=A_{21}(A_{11})^{\dagger}A_{12}=A_{21}V_{11}\Lambda_{11}^{-1}U_{11}^{\top}A_{12}$, which is a function only of blocks observed by the prediction time. The paper's central claim is that replacing those latent blocks with estimated spot volatilities and singular vectors gives a consistent predictor, and Theorem 3.1 states the element-wise error rate $O_P(\rho_m + m^{-1/4} + \phi_D/D + \sqrt{\log D/n_1} + \phi_n/n_1 + \sqrt{\log n/D})$ as $D,n_1,n,m\to\infty$. The $m^{-1/4}$ term improves on the usual nonparametric spot-volatility rate $m^{-1/8}$ because the low-rank structure lets the estimator average across the volatility surface. The same algebraic identity means no autoregressive or other dynamic volatility model is needed for the forecast.

Load-bearing premise

The load-bearing premise is that the spot-volatility estimates have errors whose conditional expectation given past information is zero, so cross terms cancel in the proof; the paper asserts this 'typically' holds but does not verify it for the two-sided pre-averaging estimator used in the numerical sections.

Editorial extensions

If this is right

  • SIP produces forecasts of the remaining intraday volatility path without an autoregressive, GARCH, HAR, or other parametric volatility model; the only structural inputs are the low-rank-plus-noise assumption and a data-driven rank $r$.
  • The convergence rate beats the $m^{-1/8}$ nonparametric spot-volatility benchmark by averaging over the volatility surface, so large historical panels $D$ and fine intraday grids $n_1$ both directly reduce prediction error.
  • Early intraday information is useful: the empirical study reports lower MSPE and QLIKE than AR, SARIMA, HAR-D, XGBoost, PC, and TIP-PCA and better 5-minute Value-at-Risk coverage even when only 10 percent of the day is observed.
  • The predictor targets the conditional expectation $E[c_{D,j}\mid F_{D,n_1}]$ of volatility, not the actual realized spot volatility $\sigma^2_{D,j}$, so forecast evaluation must compare against expected rather than noise-realized targets.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the low-rank assumption is approximately true, the same block-completion identity could be applied to option-implied volatility surfaces, multi-asset covariance matrices, or other day-by-intraday panels by swapping in the appropriate nonparametric estimator.
  • The theorem's reliance on a martingale-difference condition points to a concrete robustness question: the two-sided kernel in the numerical estimator may not satisfy that condition, and a one-sided kernel or a theorem adapted to two-sided smoothing would close the gap.
  • The assumptions suggest a ready diagnostic: inspect the singular-value gap of the estimated volatility matrix; when the gap is small relative to $\sqrt{nD}$, the low-rank completion has little signal to exploit and SIP's advantage over simpler forecasts should shrink.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper proposes a nonparametric method, the Structural Intraday-volatility Prediction (SIP) procedure, for forecasting the remaining intraday instantaneous volatility on the current trading day. The method represents the D-by-n matrix of instantaneous volatilities as a low-rank signal plus noise and predicts the missing block via the plug-in identity A22 = A21(A11)^+ A12, using estimated singular subspaces from the observed blocks. Theorem 3.1 claims an elementwise convergence rate for the predicted volatility vector, and the paper reports simulation and empirical comparisons against several benchmarks, finding that SIP achieves lower MSPE and QLIKE and better VaR coverage in the studied data.

Significance. If the theoretical claims are fully established, the paper offers a useful complement to parametric intraday volatility forecasting by exploiting low-rank structure without specifying an autoregressive or other parametric dynamics. The empirical study is extensive, covering 12 ETFs with multiple loss functions and VaR backtests, and the paper includes detailed proofs in the appendix. However, the central theoretical result is currently not justified for the estimator actually implemented in the numerical sections: the pre-averaging estimator in Eq. (4.1) uses a two-sided kernel, violating the martingale-difference condition in Assumption 3.1(iii) on which the main proof relies. There is also a technical gap in the general-r proof concerning the rotation between different singular bases. These issues make the current version unsuitable for acceptance, but they appear fixable within the manuscript's scope.

major comments (3)
  1. [Section 4, Eq. (4.1); Assumption 3.1(iii); Lemmas A.1-A.3] The pre-averaging estimator implemented in the simulations and empirical study has a two-sided kernel with bandwidth b_m = 1/n, so each spot volatility estimate uses pre-averaged returns with t_{s-1} in [τ/n - 1/n, τ/n + 1/n]. In particular, the final observed point c_{D,n1} is not measurable with respect to F_{D,n1}, and the estimation error does not satisfy the martingale-difference property E(υ_{i,j}|F_{i,t_j}) = 0 required by Assumption 3.1(iii). Lemmas A.1-A.3 repeatedly cancel cross terms using exactly this property, so Theorem 3.1 does not cover the estimator used in Sections 4 and 5. Remark 3.1's statement that the martingale condition holds when estimation uses data 'up to and including time t' does not apply to the two-sided kernel in Eq. (4.1). The authors should either implement an adapted, one-sided pre-averaging estimator in the numerical work or extend the theory to two-sided kernels.
  2. [Assumption 3.1(iv)-(vi); Lemmas A.1-A.2] Assumptions 3.1(iv) and (vi) are high-level conditions on the covariance structure of the noise matrix and on the proximity of certain population matrices, and they are used directly in the proof of the elementwise singular-vector rates in Lemma A.2. No primitive conditions on the model in Eqs. (2.1)-(2.5) or on the estimator in Eq. (4.1) are given that would verify these assumptions for the simulation DGP or the empirical setting. Since Theorem 3.1 depends on these conditions, the result is presently conditional on unverified high-level statements. Please provide sufficient primitive conditions or a direct verification for the data-generating process used in Section 4.
  3. [Lemma A.3; Theorem 3.1] For r > 1, the leading singular vectors of A11 are not the same as the leading singular vectors of the larger blocks A1• and A•1; the two sets of bases differ by an orthogonal rotation. In Lemma A.3 the proof bounds expressions involving U11 and V11 using the rates for U1• and V•1 obtained in Lemma A.2, which is only justified when r = 1. Because the SIP plug-in formula is rotation-invariant, a corrected proof should work directly with the rotated quantities that actually appear in the estimator, or the theorem should be restricted to r = 1 unless an additional assumption aligning the bases is imposed. As written, the general-r claim in Theorem 3.1 is not fully supported.
minor comments (3)
  1. [Eq. (4.1)] The uniform kernel K_{b_m}(t_{s-1} - τ/n) with bandwidth 1/n has support that extends outside the interval [0,1] for τ near 0 or n; the paper should specify how boundary points are handled.
  2. [Sections 4 and 5] The MSPE and QLIKE criteria evaluate predictions against the estimated spot volatility ĉ_{i,j} rather than the true volatility c_{i,j}; the paper should acknowledge that the reported losses include spot-volatility estimation error and explain why this does not drive the relative ranking of methods.
  3. [Simulation design and Theorem 3.1] The simulations fix m = 23,400 and n = 78 while increasing D, whereas Theorem 3.1 considers asymptotics in D, n1, n, and m jointly; the asymptotic interpretation of Figures 1 and 2 would be clearer if the paper stated how the fixed n and m relate to the theorem's conditions.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the SIP estimator is a plug-in low-rank matrix-completion predictor whose target is out-of-sample, and the derivation does not reduce to its own inputs.

full rationale

The paper's derivation chain is self-contained. Equation (2.4) (A22 = A21(A11)^† A12) is an exact algebraic identity for a rank-r matrix, and the SIP estimator replaces the unknown blocks by truncated-SVD estimates from \hatΣ11, \hatΣ12, and \hatΣ21. The target, E[c_{D,n1+j}|F_{D,n1}], is never used in constructing the estimator: \hatΣ12 contains previous days' future intraday columns, \hatΣ21 contains the current day's first n1 points, and the current day's future n2 points are not plugged into the procedure. Thus there is no fitted-input-called-prediction or self-definitional step. The low-rank structure is an assumed model, not a consequence of the estimator. The rank r is selected by an external eigenvalue-ratio rule on historical blocks (Remark 2.2), which is model selection rather than a parameter fitted to the target. Self-citations (e.g., Choi and Kim 2025) are used for motivation, DGP construction, and as a comparison method, not as load-bearing justifications of the SIP formula or its rates. The main technical concern in the manuscript—that the two-sided pre-averaging kernel in (4.1) violates the martingale-difference condition in Assumption 3.1(iii) used in Lemmas A.1–A.3—is a potential gap between the theorem and the implemented estimator, but it is a correctness/robustness issue, not circularity: the theorem's conclusion does not reduce to its assumptions by construction. The empirical study is out-of-sample against external benchmarks (AR, SARIMA, HAR-D, XGBoost, PC, TIP-PCA), and MSPE/QLIKE comparisons are based on held-out days.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a low-rank plus noise structure for the volatility matrix and on a set of high-level conditions for the spot volatility estimator and the noise. The rank is a fitted parameter in the empirical work. No new physical entities are introduced.

free parameters (3)
  • rank r = 1 (chosen by eigenvalue ratio, Remark 2.2)
    The low-rank model assumes a fixed known rank; the numerical study sets r=1 by the Ahn-Horenstein method on the same data, so the asymptotic theory with fixed r does not account for selection uncertainty.
  • kernel bandwidth b_m = 1/n (Section 4)
    Bandwidth for the pre-averaging kernel is set to 1/n rather than estimated; it controls the smoothing window for spot volatility estimation and affects the bias-variance tradeoff.
  • preaveraging window k_m = data-driven per Figueroa-Lopez and Wu (2024)
    Window size chosen by a data-driven rule; the exact selection procedure is not detailed in this paper.
assumptions (4)
  • domain assumption Interday-by-intraday volatility matrix is exactly low-rank plus noise: Σ = UΛV^T + E with rank r and λ_{r+1}=0 (Eq 2.2, Assumption 3.1(i)).
    This is the core structural assumption that makes the plug-in formula (2.4) valid; if the true matrix is not low-rank, the prediction target is misspecified.
  • ad hoc to paper Singular vectors u_k, v_k are constant vectors with incoherence bounds (Assumption 3.1(ii)).
    The phrase 'constant vectors' is stated without comment in the assumptions; it is stronger than needed and not justified from the model in Section 2.1.
  • domain assumption Spot volatility estimator error splits into martingale difference plus bias with sub-Gaussian tails and zero cross-correlations (Assumption 3.1(iii)).
    Used throughout the proof to annihilate cross terms; requires a causal (one-sided) estimator, but the implemented kernel in (4.1) appears two-sided.
  • domain assumption High-level elementwise convergence of sample second moments to a known covariance form (Assumption 3.1(iv)) and sparse idiosyncratic covariance with weak dependence (Assumptions 3.1(v)-(vi)).
    These conditions are stated as assumptions rather than derived from primitive conditions; the proof relies on them directly.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Low-Rank Structured Nonparametric Prediction of Instantaneous Volatility." pith.science (2026). https://pith.science/paper/MX43CLZE

@misc{pith2026250722173,
  author       = {Pith},
  title        = {Pith review of: Low-Rank Structured Nonparametric Prediction of Instantaneous Volatility},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MX43CLZE}},
  note         = {Machine review of arXiv:2507.22173}
}
read the original abstract

Based on It\^o semimartingale models, several studies have proposed methods for forecasting intraday volatility using high-frequency financial data. These approaches typically rely on restrictive parametric assumptions and are often vulnerable to model misspecification. To address this issue, we introduce a novel nonparametric prediction method for the future intraday instantaneous volatility process during trading hours, which leverages both previous days' data and the current day's observed intraday data. Our approach imposes an interday-by-intraday matrix representation of the instantaneous volatility, which is decomposed into a low-rank conditional expectation component and a noise matrix. To predict the future conditional expected volatility vector, we exploit this low-rank structure and propose the Structural Intraday-volatility Prediction (SIP) procedure. We establish the asymptotic properties of the SIP estimator and demonstrate its effectiveness through an out-of-sample prediction study using real high-frequency trading data.

Figures

Figures reproduced from arXiv: 2507.22173 by the authors.

Figure 1
Figure 1. MSPE×109 for the SIP, AVE, AR, SARIMA, HAR-D, XGBoost, PC, and TIP￾PCA against D with fixed ω = {0.1, 0.5}. the mean squared prediction error (MSPE) as 1 n Xn τ=n1+1 (ecD,τ − cD,τ ) 2 , where ecD,τ denotes the instantaneous volatility estimator of each method above. Finally, we calculated the sample averages of MSPEs across 500 simulations [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. MSPE×109 for the SIP, AVE, AR, SARIMA, HAR-D, XGBoost, PC, and TIP￾PCA against ω with fixed D = {50, 100}. each fixed value of ω. Figures 1 and 2 demonstrate that the SIP method consistently achieves the best perfor￾mance among the competing approaches. This may be because by incorporating information from previous intraday volatility patterns on the Dth day, SIP effectively predicts the remain￾ing future instantane… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 62 canonical work pages

  1. [1]

    Admati, A. R. and P. Pfleiderer (1988): A theory of intraday patterns: Volume and price variability, The Review of Financial Studies, 1, 3--40

    work page 1988

  2. [2]

    Ahn, S. C. and A. R. Horenstein (2013): Eigenvalue ratio test for the number of factors, Econometrica, 81, 1203--1227

    work page 2013

  3. [3]

    Fan, and D

    A \" t-Sahalia, Y., J. Fan, and D. Xiu (2010): High-frequency covariance estimates with noisy and asynchronous financial data, Journal of the American Statistical Association, 105, 1504--1517

    work page 2010

  4. [4]

    A \" t-Sahalia, Y. and J. Yu (2009): High frequency market microstructure noise estimates and liquidity measures, Annals of Applied Statistics, 3, 422--457

    work page 2009

  5. [5]

    Andersen, T. G. and T. Bollerslev (1997): Intraday periodicity and volatility persistence in financial markets, Journal of Empirical Finance, 4, 115--158

    work page 1997

  6. [6]

    Andersen, T. G., T. Bollerslev, F. X. Diebold, and P. Labys (2003): Modeling and forecasting realized volatility, Econometrica, 71, 579--625

    work page 2003

  7. [7]

    Andersen, T. G., M. Thyrsgaard, and V. Todorov (2019): Time-varying periodicity in intraday volatility, Journal of the American Statistical Association, 114, 1695--1707

    work page 2019

  8. [8]

    Bai, J. and S. Ng (2002): Determining the number of factors in approximate factor models, Econometrica, 70, 191--221

    work page 2002

Show all 64 references
  1. [9]

    --- -.1pt --- -.1pt --- (2021): Matrix completion, counterfactuals, and factor analysis of missing data, Journal of the American Statistical Association, 116, 1746--1763

    2021

  2. [10]

    Barndorff-Nielsen, O. E., P. R. Hansen, A. Lunde, and N. Shephard (2008): Designing realized kernels to measure the ex post variation of equity prices in the presence of noise, Econometrica, 76, 1481--1536

    2008

  3. [11]

    --- -.1pt --- -.1pt --- (2011): Multivariate realised kernels: consistent positive semi-definite estimators of the covariation of equity prices with noise and non-synchronous trading, Journal of Econometrics, 162, 149--169

    2011

  4. [12]

    Benjamini, Y. and Y. Hochberg (1995): Controlling the false discovery rate: a practical and powerful approach to multiple testing, Journal of the Royal Statistical Society: Series B (Methodological), 57, 289--300

    1995

  5. [13]

    Hautsch, P

    Bibinger, M., N. Hautsch, P. Malec, and M. Rei (2014): Estimating the quadratic covariation matrix from noisy observations: Local method of moments and efficiency, The Annals of Statistics, 42, 1312--1346

    2014

  6. [14]

    Boivin, J. and S. Ng (2006): Are more data always better for factor analysis? Journal of Econometrics, 132, 169--194

    2006

  7. [15]

    Cai, J.-F., E. J. Cand \`e s, and Z. Shen (2010): A singular value thresholding algorithm for matrix completion, SIAM Journal on Optimization, 20, 1956--1982

    2010

  8. [16]

    Cai, T., T. T. Cai, and A. Zhang (2016): Structured matrix completion with applications to genomic data integration, Journal of the American Statistical Association, 111, 621--633

    2016

  9. [17]

    Candes, E. and B. Recht (2012): Exact matrix completion via convex optimization, Communications of the ACM, 55, 111--119

    2012

  10. [18]

    Candes, E. J. and Y. Plan (2010): Matrix completion with noise, Proceedings of the IEEE, 98, 925--936

    2010

  11. [19]

    Chen, T. and C. Guestrin (2016): Xgboost: A scalable tree boosting system, in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785--794

    2016

  12. [20]

    Kim, and K

    Cho, J., D. Kim, and K. Rohe (2017): Asymptotic theory for estimating the singular vectors and values of a partially-observed low rank matrix with noise, Statistica Sinica, 1921--1948

    2017

  13. [21]

    Choi, S. H. and D. Kim (2025): Matrix-based prediction approach for intraday instantaneous volatility vector, Journal of Business & Economic Statistics, in press

    2025

  14. [22]

    Kinnebrock, and M

    Christensen, K., S. Kinnebrock, and M. Podolskij (2010): Pre-averaging estimators of the ex-post covariance matrix in noisy diffusion models with non-synchronous data, Journal of Econometrics, 159, 116--133

    2010

  15. [23]

    Christoffersen, P. F. (1998): Evaluating interval forecasts, International Economic Review, 841--862

    1998

  16. [24]

    (2009): A simple approximate long-memory model of realized volatility, Journal of Financial Econometrics, 7, 174--196

    Corsi, F. (2009): A simple approximate long-memory model of realized volatility, Journal of Financial Econometrics, 7, 174--196

    2009

  17. [25]

    Davis, C. and W. M. Kahan (1970): The rotation of eigenvectors by a perturbation. III, SIAM Journal on Numerical Analysis, 7, 1--46

    1970

  18. [26]

    Diebold, F. X. and R. S. Mariano (1995): Comparing predictive accuracy, Journal of Business & Economic Statistics, 13, 253--265

    1995

  19. [27]

    Engle, R. F. and S. Manganelli (2004): CAViaR: Conditional autoregressive value at risk by regression quantiles, Journal of Business & Economic Statistics, 22, 367--381

    2004

  20. [28]

    Engle, R. F. and M. E. Sokalska (2012): Forecasting intraday volatility in the US equity market. Multiplicative component GARCH, Journal of Financial Econometrics, 10, 54--83

    2012

  21. [29]

    Furger, and D

    Fan, J., A. Furger, and D. Xiu (2016): Incorporating global industrial classification standard into portfolio allocation: A simple factor-based large covariance matrix estimator with high-frequency data, Journal of Business & Economic Statistics, 34, 489--503

    2016

  22. [30]

    Fan, J. and D. Kim (2018): Robust high-dimensional volatility matrix estimation for high-frequency factor model, Journal of the American Statistical Association, 113, 1268--1283

    2018

  23. [31]

    --- -.1pt --- -.1pt --- (2019): Structured volatility matrix estimation for non-synchronized high-frequency financial data, Journal of Econometrics, 209, 61--78

    2019

  24. [32]

    Liu, and W

    Fan, J., H. Liu, and W. Wang (2018 a ): Large covariance estimation through elliptical factor models, Annals of Statistics, 46, 1383

    2018

  25. [33]

    Wang, and Y

    Fan, J., W. Wang, and Y. Zhong (2018 b ): An _ eigenvector perturbation bound and its application to robust covariance estimation, Journal of Machine Learning Research, 18, 1--42

    2018

  26. [34]

    Fan, J. and Y. Wang (2007): Multi-scale jump and volatility analysis for high-frequency financial data, Journal of the American Statistical Association, 102, 1349--1362

    2007

  27. [35]

    --- -.1pt --- -.1pt --- (2008): Spot volatility estimation for high-frequency data, Statistics and its Interface, 1, 279--288

    2008

  28. [36]

    Figueroa-L \'o pez, J. E. and B. Wu (2024): Kernel estimation of spot volatility with microstructure noise using pre-averaging, Econometric Theory, 40, 558--607

    2024

  29. [37]

    Foster, D. P. and D. B. Nelson (1996): Continuous record asymptotics for rolling sample variance estimators, Econometrica, 64, 139--174

    1996

  30. [38]

    Hansen, P. R., Z. Huang, and H. H. Shek (2012): Realized GARCH: a joint model for returns and realized measures of volatility, Journal of Applied Econometrics, 27, 877--906

    2012

  31. [39]

    Hong, H. and J. Wang (2000): Trading and returns under periodic market closures, The Journal of Finance, 55, 297--354

    2000

  32. [40]

    Jacod, J., Y. Li, P. A. Mykland, M. Podolskij, and M. Vetter (2009): Microstructure noise in the continuous case: the pre-averaging approach, Stochastic Processes and their Applications, 119, 2249--2276

    2009

  33. [41]

    Kim, D. and Y. Wang (2016 a ): Sparse PCA-based on high-dimensional It \^o processes with measurement errors, Journal of Multivariate Analysis, 152, 172--189

    2016

  34. [42]

    --- -.1pt --- -.1pt --- (2016 b ): Unified discrete-time and continuous-time models and statistical inferences for merged low-frequency and high-frequency financial data, Journal of Econometrics, 194, 220--230

    2016

  35. [43]

    Wang, and J

    Kim, D., Y. Wang, and J. Zou (2016): Asymptotic theory for large volatility matrix estimation based on high-frequency financial data, Stochastic Processes and their Applications, 126, 3527--3577

    2016

  36. [44]

    (2010): Nonparametric filtering of the realized spot volatility: A kernel-based approach, Econometric Theory, 26, 60--93

    Kristensen, D. (2010): Nonparametric filtering of the realized spot volatility: A kernel-based approach, Econometric Theory, 26, 60--93

    2010

  37. [45]

    Kupiec, P. H. (1995): Techniques for verifying the accuracy of risk measurement models, The Journal of Derivatives, 3, 73--84

    1995

  38. [46]

    Cheng, J

    Li, Q., G. Cheng, J. Fan, and Y. Wang (2018): Embracing the blessing of dimensionality in factor models, Journal of the American Statistical Association, 113, 380--389

    2018

  39. [47]

    Li, Z. M. and O. Linton (2023): Robust estimation of integrated and spot volatility, Journal of Econometrics, 105614

    2023

  40. [48]

    Mattiussi, and R

    Mancini, C., V. Mattiussi, and R. Ren \`o (2015): Spot volatility estimation using delta sequences, Finance and Stochastics, 19, 261--293

    2015

  41. [49]

    (2010): Determining the number of factors from empirical distribution of eigenvalues, The Review of Economics and Statistics, 92, 1004--1016

    Onatski, A. (2010): Determining the number of factors from empirical distribution of eigenvalues, The Review of Economics and Statistics, 92, 1004--1016

    2010

  42. [50]

    Patton, A. J. (2011): Volatility forecast comparison using imperfect volatility proxies, Journal of Econometrics, 160, 246--256

    2011

  43. [51]

    Shephard, N. and K. Sheppard (2010): Realising the future: forecasting with high-frequency-based volatility (HEAVY) models, Journal of Applied Econometrics, 25, 197--231

    2010

  44. [52]

    (2010): Financial Econometrics Notes, University of Oxford, 333--426

    Sheppard, K. (2010): Financial Econometrics Notes, University of Oxford, 333--426

    2010

  45. [53]

    Kim, and J

    Shin, M., D. Kim, and J. Fan (2023): Adaptive robust large volatility matrix estimation based on high-frequency financial data, Journal of Econometrics, 237, 105514

    2023

  46. [54]

    (2019): Nonparametric spot volatility from options, The Annals of Applied Probability, 29, 3590--3636

    Todorov, V. (2019): Nonparametric spot volatility from options, The Annals of Applied Probability, 29, 3590--3636

    2019

  47. [55]

    Todorov, V. and Y. Zhang (2023): Bias reduction in spot volatility estimation from options, Journal of Econometrics, 234, 53--81

    2023

  48. [56]

    (2010): Introduction to the non-asymptotic analysis of random matrices, arXiv preprint arXiv:1011.3027

    Vershynin, R. (2010): Introduction to the non-asymptotic analysis of random matrices, arXiv preprint arXiv:1011.3027

    2010 arXiv

  49. [57]

    Wang, W. and J. Fan (2017): Asymptotics of empirical eigenstructure for high dimensional spiked covariance, The Annals of Statistics, 45, 1342

    2017

  50. [58]

    (2010): Quasi-maximum likelihood estimation of volatility with high frequency data, Journal of Econometrics, 159, 235--250

    Xiu, D. (2010): Quasi-maximum likelihood estimation of volatility with high frequency data, Journal of Econometrics, 159, 235--250

    2010

  51. [59]

    Yan, Y. and M. J. Wainwright (2024): Entrywise inference for causal panel data: A simple and instance-optimal approach, arXiv preprint arXiv:2401.13665, 6

    2024 arXiv

  52. [60]

    Zhang, M

    Zhang, C., Y. Zhang, M. Cucuringu, and Z. Qian (2024): Volatility forecasting with machine learning and intraday commonality, Journal of Financial Econometrics, 22, 492--530

    2024

  53. [61]

    (2006): Efficient estimation of stochastic volatility using noisy observations: A multi-scale approach, Bernoulli, 12, 1019--1043

    Zhang, L. (2006): Efficient estimation of stochastic volatility using noisy observations: A multi-scale approach, Bernoulli, 12, 1019--1043

    2006

  54. [62]

    --- -.1pt --- -.1pt --- (2011): Estimating covariation: Epps effect, microstructure noise, Journal of Econometrics, 160, 33--47

    2011

  55. [63]

    Zhang, L., P. A. Mykland, and Y. A \" t-Sahalia (2005): A tale of two time scales: Determining integrated volatility with noisy high-frequency data, Journal of the American Statistical Association, 100, 1394--1411

    2005

  56. [64]

    Zu, Y. and H. P. Boswijk (2014): Estimating spot volatility with high-frequency financial data, Journal of Econometrics, 181, 117--135

    2014

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.