Pith. sign in

REVIEW 2 major objections 5 minor 74 references

A Simple Approximation to the Distribution of the Ridge Regression Estimator

T0 review · 2 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read Ridge regression's finite-sample distribution is a simple Gaussian whose bias and variance follow from the penalty and the shrinkage target.

desk verdict Theorem 1 is a clean, correct generalization of Knight–Fu to HAC errors; the tuning-rule story has an unidentifiability problem that the simulations paper over. read the letter →

arxiv 2608.02539 v1 pith:B6722UWH submitted 2026-08-03 econ.EM cs.LGmath.STstat.TH

José Luis Montiel Olea , Ryan Strong , Amilcar Velez , Zhuoheng Xu , Haomin Yu This is my paper · ORCID
classification econ.EMcs.LGmath.STstat.TH MSC 62J0762E20
keywords ridgeregressiondistributionalapproximationnonstandardasymptoticslocal-to-referencebias-variancetradeoffpredictionriskregularizationparameterselectionheteroskedasticityandautocorrelation
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

The paper aims to show that, in low-dimensional regressions with heteroskedastic or autocorrelated errors, the ridge estimator's distribution is well approximated by a Gaussian with an explicit bias and a sandwich-form variance. The approximation comes from two nonstandard asymptotics: the penalty grows proportionally to the sample size, and the true coefficients are local to the shrinkage reference vector. On top of this distributional result, the paper derives an approximation to the excess prediction risk of ridge, and then two closed-form rules for choosing the penalty: one that minimizes worst-case risk given a bound on the local parameter, and one that minimizes average risk given a prior. The paper is candid that the local parameter b is not consistently estimable, so the tuning rules require user-supplied prior or bound. If the approximation holds, it gives practitioners a way to think about ridge's bias-variance tradeoff without cross-validation or Gaussian-error assumptions.

What carries the argument

The central device is the joint rescaling lambda_n/n -> lambda and beta_n = beta_0 + b/sqrt(n) + o(1/sqrt(n)), which keeps shrinkage bias and sampling noise at the same order. The Gaussian law in Theorem 1 is the engine: its mean term -lambda(Sigma + lambda I)^{-1}b captures misspecification of the shrinkage target, and its covariance term (Sigma + lambda I)^{-1}Omega(Sigma + lambda I)^{-1} carries heteroskedasticity and autocorrelation through Omega. This approximation then feeds directly into the excess-risk function and the closed-form tuning rules.

What would settle it

Run a Monte Carlo with k fixed, x_i and epsilon_i satisfying Assumption 1, beta_n = beta_0 + b/sqrt(n), and lambda_n/n -> lambda; compare the empirical distribution of sqrt(n)(beta_hat_lambda_n - beta_n) to N(-lambda(Sigma + lambda I)^{-1}b, (Sigma + lambda I)^{-1}Omega(Sigma + lambda I)^{-1}). If the first two moments deviate systematically beyond Monte Carlo error, Theorem 1 is refuted. Setting beta fixed away from beta_0 instead should make the bias grow with sqrt(n), confirming that the local assumption is load-bearing.

Watch

Extended reading notes

Core claim

Under Assumption 1 (convergence of the design second-moment matrix Sigma and asymptotic normality of the score with covariance Omega), Theorem 1 states that if lambda_n/n converges in probability to lambda and sqrt(n)(beta_n - beta_0) converges to b, then sqrt(n)(beta_hat_lambda_n - beta_n) converges in distribution to N_k(-lambda(Sigma + lambda I_k)^{-1}b, (Sigma + lambda I_k)^{-1}Omega(Sigma + lambda I_k)^{-1}). This is the paper's central claim: in finite samples ridge is roughly Gaussian with shrinkage bias proportional to lambda and to the distance from the reference vector, and variance given by the usual sandwich form with the ridge-adjusted denominator. Theorem 2 converts this into a

Load-bearing premise

The true regression coefficients must lie within O(1/sqrt(n)) of the shrinkage reference vector beta_0; if they do not, the Gaussian approximation's bias term is wrong and the tuning rules that rely on the local parameter b lose their justification.

Editorial extensions

If this is right

  • If Theorem 1 is correct, ridge's finite-sample distribution in low-dimensional designs is a known Gaussian, so prediction intervals and risk comparisons can be made without assuming Gaussian errors or independence.
  • Heteroskedasticity and serial correlation enter only through Omega, so the approximation covers time-series and clustered data, not just i.i.d. samples.
  • The isotropic tuning formula lambda_n = n*tr(Omega_hat)/(sigma_hat_x^2 * B^2) gives an explicit ridge penalty that minimizes worst-case approximate excess risk, replacing cross-validation with a one-line calculation.
  • The paper's own DGP-3 results (k/n = 0.6) show the approximation and the minimax rule degrade away from the low-risk region and do not uniformly beat leave-one-out cross-validation, so the stated guarantees are low-dimensional.

Reading between the lines

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

  • Editorial inference: because b is unidentified, the practical content of the tuning rules is really a mapping from the user's bound or prior to a penalty; two researchers with different B or pi will get different lambda even from the same data, and nothing in the data can resolve the disagreement.
  • Editorial inference: the same local-to-reference rescaling could be applied to lasso-type estimators under Assumption 1, potentially producing analogous risk-based tuning rules; the paper suggests this extension but does not carry it out.
  • Editorial inference: the Gaussian approximation implies a simple shrinkage-bias-adjusted confidence ellipse for ridge predictions, but the paper does not develop this inferential use.
  • Editorial inference: in time-series settings Omega is the long-run variance, so the optimal lambda adapts to serial correlation; estimating Omega reliably in that setting may become the binding constraint.
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

2 major / 5 minor

Summary. The paper proposes a Gaussian approximation to the finite-sample distribution of the ridge regression estimator under high-level assumptions that allow heteroskedasticity and autocorrelation. The key asymptotic setup lets the regularization parameter grow with the sample size, λ_n/n →_p λ, and treats the true coefficient as local to the shrinkage reference vector, √n(β_n−β_0)→b. Theorem 1 derives the limiting distribution of √n(β̂_{λ_n}−β_n). Theorem 2 uses this to approximate the excess prediction risk, and Theorem 3 gives closed-form tuning rules for isotropic features: λ* = n·trace(Ω)/(σ_x²E_π[b'b]) under a prior and λ* = n·trace(Ω)/(σ_x²B²) under a norm bound. Simulation evidence compares minimax ridge with LOO-CV and OLS in three DGPs.

Significance. If the results hold, the paper provides a clean and useful generalization of Knight and Fu (2000) to heteroskedastic and autocorrelated environments and to an arbitrary reference vector. The elementary proof of Theorem 1, the transparent use of local asymptotics, and the explicit admission that b is not identified are strengths. The risk approximation and the closed-form tuning rules are new and potentially useful in low-dimensional problems. The paper is also honest about the limitation that the tuning rules require user-specified prior/bound information. However, the empirical support for the proposed tuning rules is weakened by the oracle-based calibration of B in the simulations, so the practical advantage over LOO-CV is not yet established.

major comments (2)
  1. [§3.3, Eqs. (20)–(25); Theorem 3] The tuning rules depend critically on the unidentifiable local parameter b, through either the bound B or the prior π. The paper is candid that b cannot be estimated, but it nevertheless describes the strategies as 'data-driven recommendations.' As stated, the data enter only through Ω̂ and σ̂_x²; the key tuning input B or E_π[b'b] is user-specified. More importantly, if the true b violates the bound or lies in a low-probability region of π, the selected λ can be far from the risk minimizer even though Theorem 1 still holds. The authors should add a sensitivity analysis for B and π and provide practical guidance on how to elicit these inputs, otherwise the prescriptive claim is incomplete.
  2. [§4.1–4.3, Figures 2, 5, and 8] The simulation evaluation of the minimax rule uses oracle knowledge of the true DGP to calibrate B. In DGP-1, B is set to the true ||b||=√10; in DGP-2 and DGP-3, B is the 90th percentile of √n||β_n|| computed from the same coefficient-generating scheme. LOO-CV receives no such oracle input. The reported ~20% improvement over LOO-CV in Figure 2 and the competitiveness in Figures 5 and 8 therefore reflect an oracle advantage, not a fully data-driven comparison. Please report risk as a function of B (e.g., multiples of a data-based estimate) and clarify what a practitioner must supply. This does not affect the validity of Theorem 1, but it is essential for the paper's practical claims.
minor comments (5)
  1. [§4.1, first paragraph] The error term is scalar, so 'ϵ_i ∼ N_k(0, σ²)' should be 'ϵ_i ∼ N(0, σ²)'.
  2. [§4.2–4.3] The text notes that DGP-2 and DGP-3 depart from Assumption 2 (strict stationarity) by allowing heterogeneous variances. Since Theorem 2 uses Assumption 2, this means the simulations do not exactly satisfy the theorem's conditions. The paper should state more clearly whether these simulations are intended as a robustness check outside the formal assumptions.
  3. [Figures 1, 4, and 7] The horizontal-axis notation is garbled: 'n( 1 n, 1)' should be something like '√n(β̂₁−β_{n,1})'. Please fix the typesetting.
  4. [Theorem 2, condition (iii)] The uniform integrability condition is high-level. Providing primitive sufficient conditions, such as moment bounds on the design and errors and boundedness of λ_n/n, would make the theorem easier to verify in applications.
  5. [Eq. (3) vs. Eq. (13)] The estimator is denoted with λ_n in (3) and with λ in (13), and the text later says 'with λ replaced by λ_n.' Please unify the notation to avoid confusion.

Circularity Check

1 steps flagged · score 2.0 of 10

Simulation evaluation of the minimax ridge rule uses oracle knowledge of b (B set to the truth or calibrated from the same DGP), but the theorem derivations themselves are self-contained and not circular.

  1. other [Section 4.1 (DGP-1, p.18) and Section 4.2 (DGP-2, p.22)]
    "We set B≡ ∥b∥= √10. ... For the minimax criterion, we calibrate the radius B from the same coefficient-generation scheme: for each target R2, we draw 2,000 coefficient vectors, rescale each draw to the target R2, compute √n∥βn∥, and use the 90th percentile as the baseline value of B."

    With isotropic features, the minimax rule of Theorem 3 is λ*_minimax = n trace(Ω)/(σ_x²B²). For a known b, the excess-risk minimizer over λ is the same expression with B² replaced by ∥b∥² (see the derivative (32) in the proof of Theorem 3). In DGP-1 the paper sets B exactly equal to the true ∥b∥, so the evaluated 'minimax ridge' is the oracle rule that knows b, not a fully data-driven procedure. In DGP-2/3, B is calibrated from the same coefficient-generation scheme used to draw the true β_n. The reported simulation comparisons against LOO-CV are therefore partly self-referential, although this affects the empirical evaluation rather than the theorem derivations.

full rationale

The main derivation chain is not circular. Theorem 1 is a standard Slutsky/continuous-mapping result from Assumption 1 and the stated local-to-reference condition √n(β_n−β_0)→b; the paper is explicit that b is not identified and that unbounded √n(β_n−β_0) makes the asymptotic bias diverge. Theorem 2 follows from Theorem 1 by uniform integrability and Slutsky, and Theorem 3 is closed-form algebra from the risk approximation (31)–(32). The tuning rules depend on a user-supplied bound B or prior π, which the paper states candidly; this is a limitation, not a circular derivation. The self-citations (Velez 2024; Montiel Olea et al. 2026; Shen and Xiu 2025) are contextual and not load-bearing in the proofs. The only concrete circularity concern is in the simulations, where B is set to the true coefficient norm or calibrated from the DGP's coefficient-generating scheme, making the implemented minimax rule an oracle-aided benchmark rather than a strictly data-driven competitor to LOO-CV. Because this affects the empirical demonstration and not the mathematical results, the overall circularity score is low.

Assumptions & free parameters 3 free parameters · 8 assumptions · 0 invented entities

The approximation is not parameter-free: it rests on high-level dependence assumptions, the local-to-β_0 drifting sequence, and the λ_n/n scaling. The tuning rules introduce two user-chosen hyperparameters (B or π). In simulations, an additional ad hoc n/(n−k) correction is used outside the theory. No invented entities are postulated.

free parameters (3)
  • B (radius of the local-parameter ball) = √10 in DGP-1; 90th percentile of simulated √n‖β_n‖ in DGP-2/3
    User-specified bound on ‖b‖ required by the minimax rule in Theorem 3(1). In the simulations B is set using the true DGP coefficient distribution, so it is effectively calibrated with oracle information.
  • Eπ[b^Tb] or prior hyperparameter C² = not used in simulations; user-specified
    Average-risk rule in Theorem 3(2) requires a prior mean squared norm Eπ[b^Tb], which is not data-driven.
  • Degrees-of-freedom adjustment n/(n−k) = n=500, k=50 or 300
    Ad hoc finite-sample modification applied to the variance term and tuning ratio in DGP-2/DGP-3 simulations (Section 4.2), outside the theory.
assumptions (8)
  • domain assumption Assumption 1: (1/n)Σx_ix_i^T →_p Σ and (1/√n)Σx_iε_i →_d N(0,Ω)
    High-level DGP conditions used in Theorem 1; allow HAC but are not derived from primitive conditions.
  • ad hoc to paper √n(β_n − β_0) → b (Eq. 5/14)
    Central drifting-sequence assumption; ensures the shrinkage bias is O(1) on the √n scale. Not testable and not implied by Assumption 1.
  • domain assumption λ_n/n →_p λ ∈ [0,∞) (Eq. 6/14)
    Scaling assumption for the regularization parameter; without it the bias term vanishes (Proposition 1).
  • domain assumption Strict stationarity of {(x_i, ε_i)} (Assumption 2)
    Used to define out-of-sample prediction loss with a stationary distribution P.
  • domain assumption E_P[x_iε_i]=0 and E_P[x_ix_i^T]=Σ (Theorem 2)
    Zero-score and second-moment conditions required for the excess risk formula.
  • standard math sup_n E[(Z_n^TΣZ_n)^{1+δ}] < ∞ (Theorem 2(iii))
    Technical condition to pass from distributional convergence to convergence of expectations via DasGupta (2008, Theorem 6.2).
  • domain assumption Σ = σ_x² I_k (Theorem 3)
    Restricts the closed-form tuning rules to isotropic covariates; the non-isotropic case requires numerical grid search.
  • domain assumption Eπ[b^Tb] finite and positive, or ‖b‖ ≤ B (Theorem 3)
    Decision-theoretic inputs substitute for the unestimable b; they are not data-driven.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Simple Approximation to the Distribution of the Ridge Regression Estimator." pith.science (2026). https://pith.science/paper/B6722UWH

@misc{pith2026260802539,
  author       = {Pith},
  title        = {Pith review of: A Simple Approximation to the Distribution of the Ridge Regression Estimator},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B6722UWH}},
  note         = {Machine review of arXiv:2608.02539}
}
abstract

We present a simple Gaussian approximation to the finite-sample distribution of the classical ridge regression estimator. Our approximation captures the fact that, in finite samples, the ridge regression estimator trades off bias and variance to reduce estimation and prediction error. Our approximation is based on nonstandard asymptotics where $i)$ we let the estimator's regularization parameter grow proportionally to the sample size; and $ii)$ we treat the population regression coefficients as \emph{local} to the reference vector that defines the estimator's direction of shrinkage. In contrast to other asymptotic approximations in the literature, we allow for general forms of heteroskedasticity and autocorrelation in the data generating process (at the cost of considering a low-dimensional model where the number of covariates is not allowed to grow with the sample size). We use our simple Gaussian approximation to propose two new strategies to select the regularization parameter for the ridge regression estimator. The suggested strategies select the regularization parameter to minimize either average or worst-case excess prediction risk, where risk is computed using our suggested Gaussian approximation.

Figures

Figures reproduced from arXiv: 2608.02539 by the authors.

Figure 1
Figure 1. Distribution of the first coefficient under DGP-1 [PITH_FULL_IMAGE:figures/full_fig_p019_1.png] view at source ↗
Figure 2
Figure 2. Relative excess prediction risk under DGP-1 [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗
Figure 3
Figure 3. Risk curves under DGP-1 We also compare the scaled regularization parameter (λbn/n) used by the LOO-CV and minimax ridge estimators [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗
Figures from the paper (13 more)
Figure 4
Figure 4. Figure 4: Distribution of the first coefficient under DGP-2 [PITH_FULL_IMAGE:figures/full_fig_p023_4.png]
Figure 5
Figure 5. Figure 5: Relative excess prediction risk under DGP-2, LOO-CV and minimax [PITH_FULL_IMAGE:figures/full_fig_p024_5.png]
Figure 6
Figure 6. Figure 6: Risk curves under DGP-2 4.3 DGP-3 We next consider the high-dimensional version of the weak-signal design based on Section 3.1 of Shen and Xiu (2025). Our asymptotic theory focused on models in which the number of covariates is small relative to the sample size. Thus, …
Figure 7
Figure 7. Figure 7: Distribution of the first coefficient under DGP-3 [PITH_FULL_IMAGE:figures/full_fig_p026_7.png]
Figure 8
Figure 8. Figure 8: Relative excess prediction risk under DGP-3, LOO-CV and minimax [PITH_FULL_IMAGE:figures/full_fig_p027_8.png]
Figure 9
Figure 9. Figure 9: Risk curves under DGP-3 References Abadie, A. and M. Kasy (2019): “Choosing among Regularized Estimators in Empirical Economics: The Risk of Machine Learning,” The Review of Economics and Statistics, 101, 743–762. Andrews, I. and A. Mikusheva (2022): “Optimal decision …
Figure 10
Figure 10. Figure 10: Scaled regularization parameter (λbn/n) used by the LOO-CV and minimax ridge estimators under DGP-1 B.2.2 DGP-2 This appendix records diagnostic information for the covariance matrix and coefficient vec￾tors used in DGP-2. These diagnostics are useful because the desi…
Figure 11
Figure 11. Figure 11: Eigenvalues of Σ2 under DGP-2 [PITH_FULL_IMAGE:figures/full_fig_p037_11.png]
Figure 12
Figure 12. Figure 12: Coefficient vectors under DGP-2 increasing the target R2 changes the magnitude of the coefficient vector while preserving the same coefficient pattern [PITH_FULL_IMAGE:figures/full_fig_p038_12.png]
Figure 13
Figure 13. Figure 13: Selected tuning ratios under DGP-2 because the diagonal entries of Σ1 are equal to one. Since the errors are independent of the regressors, mean zero, and have fixed variances σ 2 ϵ,i, the corresponding score variance matrix is 1 n Xn i=1 E[xix ⊤ i ϵ 2 i ] = 1 n Xn i=…
Figure 14
Figure 14. Figure 14: Eigenvalues of Σ2 under DGP-3 0.15 0.10 0.05 0.00 0.05 0.10 0.15 n, j R 2 = 5%; n = 0.2252 0.15 0.10 0.05 0.00 0.05 0.10 0.15 n, j R 2 = 20%; n = 0.4908 0 50 100 150 200 250 300 Coefficient index 0.15 0.10 0.05 0.00 0.05 0.10 0.15 n, j R 2 = 50%; n = 0.9815 [PITH_FUL…
Figure 15
Figure 15. Figure 15: Coefficient vectors under DGP-3 and minimax ridge under DGP-3. Relative to DGP-2, the LOO-CV choices are substantially more dispersed, especially when R2 is small. This dispersion is consistent with the main-text 40 [PITH_FULL_IMAGE:figures/full_fig_p040_15.png]
Figure 16
Figure 16. Figure 16: Selected tuning ratios under DGP-3 finding that, in the high-dimensional design, LOO-CV and minimax ridge have comparable average risk but can choose very different penalty values in individual samples. 41 [PITH_FULL_IMAGE:figures/full_fig_p041_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 3 canonical work pages

  1. [1]

    Abadie, A. and M. Kasy (2019): Choosing among Regularized Estimators in Empirical Economics: The Risk of Machine Learning, The Review of Economics and Statistics, 101, 743--762

    2019

  2. [2]

    Andrews, I. and A. Mikusheva (2022): Optimal decision rules for weak GMM, Econometrica, 90, 715--748

    2022

  3. [3]

    Anup, A. K. and G. Maddala (1984): Ridge estimators for distributed lag models, Communications in Statistics-Theory and Methods, 13, 217--225

    1984

  4. [5]

    Brockwell, P. J. and R. A. Davis (2013): Time series: theory and methods, Springer Science & Business Media

    2013

  5. [6]

    Huang, and G

    Cambanis, S., S. Huang, and G. Simons (1981): On the theory of elliptically contoured distributions, Journal of Multivariate Analysis, 11, 368--385

    1981

  6. [7]

    Cattaneo, M. D., M. Jansson, and W. K. Newey (2018): Inference in linear regression models with many covariates and heteroscedasticity, Journal of the American Statistical Association, 113, 1350--1361

    2018

  7. [8]

    (2008): Asymptotic Theory of Statistics and Probability, Springer Verlag

    DasGupta, A. (2008): Asymptotic Theory of Statistics and Probability, Springer Verlag

    2008

  8. [9]

    Dobriban, E. and S. Wager (2018): High-dimensional asymptotics of prediction: Ridge regression and classification, The Annals of Statistics, 46, 247--279

    2018

Show all 74 references
  1. [10]

    Dou, L. and U. K. M \"u ller (2021): Generalized Local-to-Unity Models, Econometrica, 89, 1825--1854

    2021

  2. [11]

    Ferguson, T. S. (1967): Mathematical Statistics: A Decision Theoretic Approach, New York: Academic Press

    1967

  3. [12]

    Hastie, and R

    Friedman, J., T. Hastie, and R. Tibshirani (2017): The elements of statistical learning: data mining, inference and prediction, vol. 1 of Series in Statistics, New York: Springer, second edition ed

    2017

  4. [13]

    Gibbons, D. G. (1981): A Simulation Study of Some Ridge Estimators, Journal of the American Statistical Association, 76, 131--139

    1981

  5. [14]

    (2022): Econometrics, Princeton University Press

    Hansen, B. (2022): Econometrics, Princeton University Press

    2022

  6. [15]

    Hansen, B. E. (2016): Efficient shrinkage in parametric models, Journal of Econometrics, 190, 115--132

    2016

  7. [16]

    Montanari, S

    Hastie, T., A. Montanari, S. Rosset, and R. J. Tibshirani (2022): Surprises in high-dimensional ridgeless least squares interpolation, The Annals of Statistics, 50, 949--986

    2022

  8. [17]

    Hirano, K. and J. H. Wright (2017): Forecasting with model uncertainty: Representations and risk reduction, Econometrica, 85, 617--643

    2017

  9. [18]

    Hoerl, A. E. (1962): Application of ridge analysis to regression problems, Chemical Engineering Progress, 58, 54--59

    1962

  10. [19]

    Hoerl, A. E. and R. W. Kennard (1970): Ridge regression: Biased estimation for nonorthogonal problems, Technometrics, 12, 55--67

    1970

  11. [20]

    Hoerl, R. W. (2020): Ridge regression: a historical context, Technometrics, 62, 420--425

    2020

  12. [21]

    Hsu, D., S. M. Kakade, and T. Zhang (2012): Random design analysis of ridge regression, in Conference on learning theory, JMLR Workshop and Conference Proceedings, 9--1

    2012

  13. [22]

    Knight, K. and W. Fu (2000): Asymptotics for lasso-type estimators, Annals of statistics, 1356--1378

    2000

  14. [23]

    Kock, A. B., R. S. Pedersen, and J. R.-V. S rensen (2026): Data-driven tuning parameter selection for high-dimensional vector autoregressions, Journal of the American Statistical Association, 121, 289--299

    2026

  15. [24]

    Zheng, and X

    Liu, X., S. Zheng, and X. Feng (2020): Estimation of error variance via ridge regression, Biometrika, 107, 481--488

    2020

  16. [25]

    Montiel Olea, J. L., C. Rush, A. Velez, and J. Wiesel (2026): The distributionally robust prediction error of the LASSO and related estimators, The Annals of Statistics, 54, 1006--1027

    2026

  17. [26]

    (2022): Exact minimax risk for linear least squares, and the lower tail of sample covariance matrices, The Annals of Statistics, 50, 2157--2178

    Mourtada, J. (2022): Exact minimax risk for linear least squares, and the lower tail of sample covariance matrices, The Annals of Statistics, 50, 2157--2178

    2022

  18. [27]

    Mourtada, J. and L. Rosasco (2022): An elementary analysis of ridge regression with random design, Comptes Rendus. Math \'e matique , 360, 1055--1063

    2022

  19. [29]

    Patil, P., Y. Wei, A. Rinaldo, and R. Tibshirani (2021): Uniform consistency of cross-validation estimators for high-dimensional ridge regression, in International conference on artificial intelligence and statistics, PMLR, 3178--3186

    2021

  20. [30]

    Phillips, P. C. (1988): Regression theory for near-integrated time series, Econometrica: Journal of the Econometric Society, 1021--1043

    1988

  21. [31]

    Powell, J. L. (2017): Identification and Asymptotic Approximations: Three Examples of Progress in Econometric Theory, Journal of Economic Perspectives, 31, 107--124

    2017

  22. [32]

    Shen, Z. and D. Xiu (2025): Can Machines Learn Weak Signals? Working Paper 33421, National Bureau of Economic Research, Cambridge, MA

    2025

  23. [33]

    Staiger, D. and J. H. Stock (1997): Instrumental Variables Regression with Weak Instruments, Econometrica, 65, 557--586

    1997

  24. [34]

    Swindel, B. F. (1976): Good ridge estimators based on prior information, Communications in Statistics-Theory and Methods, 5, 1065--1075

    1976

  25. [36]

    Golub and Michael Heath and Grace Wahba , journal =

    Gene H. Golub and Michael Heath and Grace Wahba , journal =. Generalized Cross-Validation as a Method for Choosing a Good Ridge Parameter , urldate =

  26. [37]

    2025 , eprint=

    Risk and cross validation in ridge regression with correlated samples , author=. 2025 , eprint=

    2025

  27. [38]

    and Barajas-Ramirez, Eduardo A

    Montesinos-López, Osval A. and Barajas-Ramirez, Eduardo A. and Montesinos-López, Abelardo and Lecumberry, Federico and Fariello, María Inés and Montesinos-López, José Cricelio and Ramirez Alcaraz, Juan Manuel and Crossa, José and Howard, Reka , TITLE =. Genes , VOLUME =. 2025 ...

    2025

  28. [39]

    Teaching Econometrics , chapter=

    Katherine Hauck and Tiemen Woutersen , title=. Teaching Econometrics , chapter=. 2026 , month=. doi:10.1007/978-3-031-97942-2_10 , url=

    2026 doi

  29. [40]

    , publisher =

    DasGupta, A. , publisher =. Asymptotic Theory of Statistics and Probability , year =

  30. [41]

    Deep Learning , Year =

    Goodfellow, Ian and Bengio, Yoshua and Courville, Aaron , Publisher =. Deep Learning , Year =

  31. [42]

    The Annals of Statistics , volume=

    High-dimensional asymptotics of prediction: Ridge regression and classification , author=. The Annals of Statistics , volume=. 2018 , publisher=

    2018

  32. [43]

    The Annals of Statistics , volume=

    Surprises in high-dimensional ridgeless least squares interpolation , author=. The Annals of Statistics , volume=. 2022 , publisher=

    2022

  33. [44]

    Comptes Rendus

    An elementary analysis of ridge regression with random design , author=. Comptes Rendus. Math

  34. [45]

    arXiv preprint arXiv:2408.04607 , year=

    Risk and cross validation in ridge regression with correlated samples , author=. arXiv preprint arXiv:2408.04607 , year=

    arXiv

  35. [46]

    2022 , publisher=

    Econometrics , author=. 2022 , publisher=

    2022

  36. [47]

    The elements of statistical learning: data mining, inference and prediction , volume =

    Friedman, Jerome and Hastie, Trevor and Tibshirani, Robert , edition =. The elements of statistical learning: data mining, inference and prediction , volume =

  37. [48]

    Communications in Statistics-Theory and Methods , volume=

    Good ridge estimators based on prior information , author=. Communications in Statistics-Theory and Methods , volume=. 1976 , publisher=

    1976

  38. [49]

    Communications in Statistics-Theory and Methods , volume=

    Ridge estimators for distributed lag models , author=. Communications in Statistics-Theory and Methods , volume=. 1984 , publisher=

    1984

  39. [50]

    Annals of statistics , pages=

    Asymptotics for lasso-type estimators , author=. Annals of statistics , pages=. 2000 , publisher=

    2000

  40. [51]

    Journal of Economic Perspectives , volume =

    Identification and Asymptotic Approximations: Three Examples of Progress in Econometric Theory , author =. Journal of Economic Perspectives , volume =. 2017 , doi =

    2017

  41. [52]

    Stock , title =

    Douglas Staiger and James H. Stock , title =. Econometrica , volume =. 1997 , doi =

    1997

  42. [53]

    2025 , month = jan, address =

    Zhouyu Shen and Dacheng Xiu , title =. 2025 , month = jan, address =. doi:10.3386/w33421 , url =

    2025 doi

  43. [54]

    Econometrica: Journal of the Econometric Society , pages=

    Regression theory for near-integrated time series , author=. Econometrica: Journal of the Econometric Society , pages=. 1988 , publisher=

    1988

  44. [55]

    Econometrica , volume=

    Optimal decision rules for weak GMM , author=. Econometrica , volume=. 2022 , publisher=

    2022

  45. [56]

    Econometrica , volume=

    Generalized Local-to-Unity Models , author=. Econometrica , volume=. 2021 , publisher=

    2021

  46. [57]

    Time series: theory and methods , year =

    Brockwell, Peter J and Davis, Richard A , publisher =. Time series: theory and methods , year =

  47. [58]

    The Annals of Statistics , volume=

    Exact minimax risk for linear least squares, and the lower tail of sample covariance matrices , author=. The Annals of Statistics , volume=. 2022 , publisher=

    2022

  48. [59]

    The Review of Economics and Statistics , volume =

    Abadie, Alberto and Kasy, Maximilian , title =. The Review of Economics and Statistics , volume =. 2019 , month =. doi:10.1162/rest_a_00812 , url =

    2019 doi

  49. [60]

    2026 , eprint=

    From Cross-Validation to SURE: Asymptotic Risk of Tuned Regularized Estimators , author=. 2026 , eprint=

    2026

  50. [61]

    On the theory of elliptically contoured distributions , journal =

    Stamatis Cambanis and Steel Huang and Gordon Simons , keywords =. On the theory of elliptically contoured distributions , journal =. 1981 , issn =. doi:https://doi.org/10.1016/0047-259X(81)90082-8 , url =

    1981 doi

  51. [62]

    Current Population Survey, 2009 Annual Social and Economic Supplement (ASEC) [Machine-readable data file and technical documentation] , year =

    2009

  52. [63]

    International conference on artificial intelligence and statistics , pages=

    Uniform consistency of cross-validation estimators for high-dimensional ridge regression , author=. International conference on artificial intelligence and statistics , pages=. 2021 , organization=

    2021

  53. [64]

    Ferguson , title =

    Thomas S. Ferguson , title =. 1967 , isbn =

    1967

  54. [65]

    Chemical Engineering Progress , volume=

    Application of ridge analysis to regression problems , author=. Chemical Engineering Progress , volume=

  55. [66]

    Technometrics , volume=

    Ridge regression: Biased estimation for nonorthogonal problems , author=. Technometrics , volume=. 1970 , publisher=

    1970

  56. [67]

    Technometrics , volume=

    Ridge regression: a historical context , author=. Technometrics , volume=. 2020 , publisher=

    2020

  57. [68]

    arXiv preprint arXiv:2411.01864 , year=

    On the asymptotic properties of debiased machine learning estimators , author=. arXiv preprint arXiv:2411.01864 , year=

    arXiv

  58. [69]

    Conference on learning theory , pages=

    Random design analysis of ridge regression , author=. Conference on learning theory , pages=. 2012 , organization=

    2012

  59. [70]

    A Simulation Study of Some Ridge Estimators , urldate =

    Diane Galarneau Gibbons , journal =. A Simulation Study of Some Ridge Estimators , urldate =

  60. [71]

    arXiv preprint arXiv:2404.01233 , year=

    Optimal ridge regularization for out-of-distribution prediction , author=. arXiv preprint arXiv:2404.01233 , year=

    arXiv

  61. [72]

    The Annals of Statistics , volume=

    The distributionally robust prediction error of the LASSO and related estimators , author=. The Annals of Statistics , volume=. 2026 , publisher=

    2026

  62. [73]

    Journal of Econometrics , volume=

    Efficient shrinkage in parametric models , author=. Journal of Econometrics , volume=. 2016 , publisher=

    2016

  63. [74]

    Biometrika , volume=

    Estimation of error variance via ridge regression , author=. Biometrika , volume=. 2020 , publisher=

    2020

  64. [75]

    Journal of the American Statistical Association , volume=

    Inference in linear regression models with many covariates and heteroscedasticity , author=. Journal of the American Statistical Association , volume=. 2018 , publisher=

    2018

  65. [76]

    Journal of the American Statistical Association , volume=

    Data-driven tuning parameter selection for high-dimensional vector autoregressions , author=. Journal of the American Statistical Association , volume=. 2026 , publisher=

    2026

  66. [77]

    Econometrica , volume=

    Forecasting with model uncertainty: Representations and risk reduction , author=. Econometrica , volume=. 2017 , publisher=

    2017

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.