The reviewed record of science sign in
Pith

arxiv: 2607.05350 · v1 · pith:7YBCXVYK · submitted 2026-07-06 · econ.EM

Approximate Minimax Estimation of a Bounded Normal Mean via Stochastic Mirror Ascent

Jos\'e Luis Montiel Olea , Ekaterina Zubova This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-07 15:27 UTCglm-5.2pith:7YBCXVYKrecord.jsonopen to challenge →

classification econ.EM
keywords Bounded Normal Meanminimax estimationstochastic mirror ascentleast-favorable distributionstatistical decision theoryimpulse response estimationLipschitz risk function
0
0 comments X

The pith

Algorithm computes near-minimax estimator for bounded normal mean

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

This paper presents a computational method for finding an approximately minimax estimator of a bounded normal mean — the classical problem of estimating the mean of a normal distribution when that mean is known to lie in a bounded interval. The authors use stochastic mirror ascent, an iterative optimization algorithm that exploits the geometry of the probability simplex, to find an approximately least-favorable distribution: the prior distribution that an adversary would choose to make the statistician's job hardest. The Bayes estimator under this approximate prior then serves as the approximately minimax estimator. The paper provides four main theorems. Theorem 1 shows that the algorithm, with explicitly specified step size, grid size, and iteration count, outputs a distribution whose Bayes risk is within a user-controlled tolerance of the maximin value, with high probability. Theorems 2 through 4 establish that as the tolerance shrinks, the approximate least-favorable distribution converges in Wasserstein distance to the true one, and the corresponding Bayes estimator converges uniformly over compact sets to the exact minimax estimator. A key technical ingredient is a new Lipschitz continuity result for the risk function (Lemma 1), which controls the discretization error introduced by restricting optimization to a finite grid. Simulations show risk improvements of 6–18% over the best linear minimax estimator, which is known to be at most 20% suboptimal. The method is applied to the practical problem of combining local projection and vector autoregression estimates of macroeconomic impulse responses, where the bounded normal mean problem arises naturally from bounding the bias of the VAR estimator.

Core claim

The central discovery is that the least-favorable distribution of the bounded normal mean problem — which has no closed-form solution for moderate bounds — can be approximated to arbitrary precision by running stochastic mirror ascent on a discretized version of the maximin problem, and that both sources of error (discretization and optimization) can be explicitly bounded and controlled. The risk function's Lipschitz continuity in the parameter, with a constant that is uniform across all decision rules, is the key property that makes the discretization error tractable. This, combined with the well-understood convergence theory of mirror ascent on the simplex, yields a procedure with end-to-端

What carries the argument

Stochastic mirror ascent with negative-entropy mirror map on the probability simplex; Lipschitz continuity of the risk function R(d,θ) in θ with constant 3m²+4m; discretization of the parameter space to a finite grid; multiplicative weights update for prior distribution; Monte Carlo estimation of supergradients using a single draw per grid point per iteration.

If this is right

  • The algorithm makes minimax estimation practically accessible for the bounded normal mean problem, which serves as a building block in nonparametric estimation and robust inference beyond econometrics.
  • The framework of controlling discretization error via Lipschitz continuity of the risk function, plus optimization error via mirror ascent theory, could be applied to other statistical decision problems where the least-favorable distribution lacks a closed form.
  • The application to combining LP and VAR estimators suggests that nonlinear bias correction of VAR estimates can systematically outperform linear shrinkage, with the degree of improvement tied to the tightness of the assumed bias bound.
  • The gap between the theoretically required iteration count (millions for moderate m) and the practically achieved accuracy suggests that adaptive step-size methods could substantially reduce computational cost while maintaining guarantees.

Where Pith is reading between the lines

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

  • The Lipschitz constant 3m²+4m used in the analysis is derived from a total variation bound on normal distributions that is known to be improvable by a factor of roughly √(2/π); adopting the sharper bound would reduce both the required grid size and iteration count.
  • The observation that empirical supergradient magnitudes are 2–5x smaller than the worst-case Lipschitz bound suggests that data-adaptive step sizes (as in AdaGrad-style methods) could close much of the gap between theoretical and practical iteration counts, though extending high-probability guarantees to data-dependent step sizes would require new concentration arguments.
  • The restriction to an ex ante fixed grid could be lifted by algorithms that incrementally add support points, potentially exploiting the known finite-support property of the true least-favorable distribution to reduce dimensionality.

Load-bearing premise

The convergence guarantee requires a number of iterations that scales as m⁴ ln(m), reaching roughly 2.3 million for m=3, driven by a worst-case Lipschitz constant M=4m² that the authors' own simulations suggest overstates the required computational effort by a factor of 2–5. The authors note this bound is sufficient but not necessary, and that far fewer iterations suffice in practice.

What would settle it

If the risk function failed to be Lipschitz continuous in θ uniformly over decision rules, the discretization error would not be controllable by simply refining the grid, and the entire approach of reducing the infinite-dimensional maximin problem to a finite-dimensional one would break down. Additionally, if the empirical supergradient magnitudes were systematically larger than the worst-case bound suggests (rather than smaller, as observed), the practical performance would match the pessimistic theoretical iteration counts and the method would be computationally infeasible for moderate m.

Figures

Figures reproduced from arXiv: 2607.05350 by Ekaterina Zubova, Jos\'e Luis Montiel Olea.

Figure 1
Figure 1. Figure 1: Number of iterations and grid size as functions of [PITH_FULL_IMAGE:figures/full_fig_p023_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Computation time and ϵ(m). 4.2 Improvement of the approximately minimax estimator rela￾tive to the minimax linear estimator [PITH_FULL_IMAGE:figures/full_fig_p024_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: is thus graphical evidence of the success of our stochastic mirror ascent algorithm [PITH_FULL_IMAGE:figures/full_fig_p025_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Lower bound on percentage improvement in worst-case risk of [PITH_FULL_IMAGE:figures/full_fig_p026_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of different estimators for the Bounded Normal Mean problem. [PITH_FULL_IMAGE:figures/full_fig_p027_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Required iterations vs. completed within five-minute limit for [PITH_FULL_IMAGE:figures/full_fig_p027_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Risk of dπϵ vs. linear minimax (five-minute limit) [PITH_FULL_IMAGE:figures/full_fig_p028_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Risk improvement of proposed estimator vs. linear (five- and fifteen-minute limits). [PITH_FULL_IMAGE:figures/full_fig_p028_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: LP vs. VAR impulse responses in four applications. [PITH_FULL_IMAGE:figures/full_fig_p037_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Hausman statistic across horizons. For each application, we compute the approximate rule d ∗ BNM-m in the corresponding minimax Bounded Normal Mean problem using stochastic mirror ascent. The bias correction of the VAR estimator is nonlinear in ∆h and is given by σ∆h d ∗ BNM-m(∆h) [PITH_FULL_IMAGE:figures/full_fig_p038_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Bias correction. To benchmark the nonlinear bias-corrected VAR, we compare it with the minimax￾optimal linear combination of LP and VAR. That is, a linear combination of VAR and LP estimators, with weights chosen to minimize the worst-case risk. As shown in Montiel Olea et al. [2026, p. 40], the worst-case risk for such a linear combination is given by (1 − ω) 2m2 (σ 2 LP − σ 2 VAR) + σ 2 VAR + ω 2 (σ 2 L… view at source ↗
Figure 12
Figure 12. Figure 12: Comparison of VAR, LP and proposed bias-corrected VAR estimators. [PITH_FULL_IMAGE:figures/full_fig_p040_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Risk improvement over linear combination. [PITH_FULL_IMAGE:figures/full_fig_p041_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Comparison of theoretical and empirical Lipschitz constants. [PITH_FULL_IMAGE:figures/full_fig_p070_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Comparison of the number of iterations using the worst-case bound [PITH_FULL_IMAGE:figures/full_fig_p071_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Worst-case risk of dπϵ (certified bound) vs. alternative estimators and analytical bounds. the procedure where the exact answer is known [PITH_FULL_IMAGE:figures/full_fig_p072_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: dπϵ vs. linear vs. true minimax. D.5 Stability of the risk evaluation under Monte Carlo random￾ness One concern with the estimator dπϵ is that—due to randomness introduced by the stochas￾tic mirror ascent procedure—different runs of the algorithm could generate very different estimators. One way to assuage this concern is to report the behavior of the risk function across different seeds [PITH_FULL_IMAGE… view at source ↗
Figure 18
Figure 18. Figure 18: Risk of dπϵ for m = 2. 72 [PITH_FULL_IMAGE:figures/full_fig_p073_18.png] view at source ↗
read the original abstract

This paper presents a computational approach to find an approximately minimax estimator for the classical Bounded Normal Mean problem. The suggested procedure is the Bayes estimator corresponding to an approximately least-favorable distribution obtained from a stochastic mirror ascent routine for concave maximization. The paper shows that both the approximately least-favorable distribution and the approximately minimax estimator are indeed close (in a sense we make precise) to their desired targets. Simulation evidence suggests that the approximately minimax estimator can yield, with a reasonable amount of compute, risk improvements from 6% to almost 18% relative to the minimax linear estimator (which is known to admit a maximal improvement of 20%). The approximately minimax estimator is then applied to the problem of how to best aggregate the information contained in local projections and vector autoregressions to estimate an impulse response coefficient.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 7 minor

Summary. This paper proposes a computational approach to constructing an approximately minimax estimator for the classical Bounded Normal Mean (BNM) problem. The authors use stochastic mirror ascent with a negative-entropy mirror map to approximately solve the concave maximization problem of finding a least-favorable distribution over a discretized parameter space. The paper provides four main theorems: Theorem 1 gives a high-probability bound on the approximation quality of the least-favorable distribution as a function of the step size, number of iterations, and grid size; Theorem 2 establishes convergence in 1-Wasserstein distance to the true least-favorable distribution; Theorem 3 shows the resulting Bayes estimators form a stochastic minimax sequence; and Theorem 4 proves uniform convergence over compact sets to the exact minimax estimator. Simulations demonstrate 6–18% risk improvements over the minimax linear estimator for m in {1,...,4}, and the method is applied to aggregating LP and VAR impulse response estimates in four macroeconomic applications.

Significance. The paper makes a genuine contribution by providing explicit, finite-sample theoretical guarantees for a computational approach to the BNM problem—a classical problem where exact minimax estimators are unavailable for general m. The combination of a new Lipschitz continuity result for the risk function (Lemma 1), a clean discretization bound (Lemma 2), and the stochastic mirror ascent convergence guarantee (Theorem 1) with explicit parameter recommendations is a strength. The extension to stochastic minimax sequences (Definition 1, Theorem 3) and uniform convergence over compacts (Theorem 4) are well-motivated and correctly derived. The application to LP–VAR aggregation via the Armstrong–Kline–Sun framework connects the theoretical results to a problem of active interest in macroeconometrics. The authors are transparent about the conservativeness of their bounds and the gap between theoretical iteration counts and practical performance, which is commendable.

major comments (2)
  1. Section 4.4, Figures 6–8: The paper reports risk improvements of 6–18% for m in {1,...,4}, but for m >= 2.2 the wall-clock time limits (5 or 15 minutes) bind before the Theorem 1 iteration count J(epsilon) is reached. This means the formal epsilon-minimax certification from Theorem 1 does not apply to a subset of the headline simulation results. The authors acknowledge this (Remark 2 in Section 3, and the discussion in Section 4.4), but the abstract and introduction state the 6–18% improvement range without qualifying that some of these results lack formal certification. I recommend that the abstract and the simulation summary explicitly distinguish between results where the Theorem 1 guarantee applies (m approximately <= 2.2) and results where only empirical performance is demonstrated. This is a presentation issue that affects how readers interpret the central empirical claim, but itis
  2. Section 4.4 and Figures 7–8: When the wall-clock limit binds and the algorithm does not complete J(epsilon) iterations, the output distribution pi_epsilon is not guaranteed to be epsilon-least-favorable. The paper reports a 'certified upper bound' on the worst-case risk in Figure 3 (using the Lipschitz property from Lemma 1), but it is unclear whether this certification strategy is still valid when the algorithm has been stopped early. The Lipschitz-based certification in Appendix D.3 bounds the gap between grid-evaluated risk and the true supremum, but it does not address the optimization error from incomplete iterations. The authors should clarify that the certified upper bound in Figure 3 applies only to the discretization/Monte Carlo gap, not to the optimization error, and that for time-limited runs the optimization error is uncontrolled. This distinction is important for readers who
minor comments (7)
  1. The notation v*(m) is used for the maximin value (Equation 6) and v-bar*(m) for the minimax value (Equation 2). These are shown to be equal (Equation 7), but the dual use of v* and v-bar* throughout the paper can cause confusion. Consider unifying the notation after Equation 7 is established, or adding a remark that they coincide henceforth.
  2. In Algorithm 1, Step 6, the Bayes decision rule d_pi_j(y) is written as a ratio involving theta_i and phi(y;theta_i). The formula is correct but could benefit from a brief inline explanation that this is the posterior mean under prior pi_j, to help readers less familiar with the BNM literature.
  3. Figure 1 shows J(m) and I(m) as functions of m, but the y-axis scale for J(m) is not clearly labeled (it appears to be in units of 10^6 for larger m). Adding explicit axis labels would improve readability.
  4. The reference to 'Proposition 4.2 in Ben-Tal, Margalit, and Nemirovski [2001]' on page 2 is cited to support the claim that mirror ascent is 'essentially the best first-order iterative algorithm' for this problem. The claim is strong; the authors should verify that the proposition's setting (ordered subsets mirror descent for tomography) directly supports this optimality claim for the BNM simplex problem, or soften the language.
  5. In Section 5.4, the values of m across applications (Table 1) range from 1.52 to 1.64. These are below 2.2, so the Theorem 1 guarantee should apply. It would be useful to explicitly state that the formal certification applies to all four applications, as this strengthens the applied results.
  6. The paper mentions (footnote 3) that off-the-shelf methods like fmincon might perform comparably but lack theoretical guarantees. A brief comparison with such methods in the simulation section would strengthen the case for the proposed algorithm, even if only to confirm that the mirror ascent approach is competitive in practice.
  7. Typo on page 28, Section 4.5: 'factors of approximately 2, 3.5, and 5' should clarify these are reduction factors in the number of iterations, not in runtime. Also, Figure 15 in Appendix D is referenced but the factors (2x, 3.5x, 5x) should be explicitly tied to the corresponding m values (1, 1.6, 2).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for a careful and constructive report. We are grateful for the positive assessment of the paper's theoretical contributions and the recommendation of minor revision. Both major comments raise legitimate and related points about the presentation of simulation results when the wall-clock time limit binds before the Theorem 1 iteration count is reached. We address each comment below.

read point-by-point responses

  1. Referee: Section 4.4, Figures 6-8: The paper reports risk improvements of 6-18% for m in {1,...,4}, but for m >= 2.2 the wall-clock time limits bind before the Theorem 1 iteration count J(epsilon) is reached. This means the formal epsilon-minimax certification from Theorem 1 does not apply to a subset of the headline simulation results. The authors acknowledge this, but the abstract and introduction state the 6-18% improvement range without qualifying that some of these results lack formal certification. The referee recommends that the abstract and simulation summary explicitly distinguish between results where the Theorem 1 guarantee applies (m approximately <= 2.2) and results where only empirical performance is demonstrated.

    Authors: The referee is correct that the abstract and introduction do not currently distinguish between results with formal Theorem 1 certification and results where only empirical performance is demonstrated. We agree this is an important presentation issue and will revise accordingly. Specifically, we will modify the abstract to note that the 6-18% range includes both certified results (for smaller m, where the algorithm completes the theoretically required number of iterations) and empirically demonstrated improvements (for larger m, where wall-clock time limits bind before J(epsilon) is reached). We will add a similar clarification in the introduction and in the simulation summary in Section 4.4. This will make transparent which subset of results carries the formal epsilon-minimax guarantee from Theorem 1 and which subset reports only empirical performance. revision: yes

  2. Referee: Section 4.4 and Figures 7-8: When the wall-clock limit binds and the algorithm does not complete J(epsilon) iterations, the output distribution pi_epsilon is not guaranteed to be epsilon-least-favorable. The paper reports a 'certified upper bound' on the worst-case risk in Figure 3 (using the Lipschitz property from Lemma 1), but it is unclear whether this certification strategy is still valid when the algorithm has been stopped early. The Lipschitz-based certification in Appendix D.3 bounds the gap between grid-evaluated risk and the true supremum, but it does not address the optimization error from incomplete iterations. The authors should clarify that the certified upper bound in Figure 3 applies only to the discretization/Monte Carlo gap, not to the optimization error, and that for time-limited runs the optimization error is uncontrolled.

    Authors: The referee is correct on the substance. The Lipschitz-based certification in Appendix D.3 addresses only the gap between the worst-case risk evaluated on the finite grid and the true supremum over the continuous parameter space [−m, m]. It does not address the optimization error that arises when the algorithm is stopped before completing J(epsilon) iterations. We will revise the paper to make this distinction explicit. Specifically, we will add clarifying language in Section 4.4 and in Appendix D.3 stating that: (i) the certified upper bound in Figure 3 (and the corresponding bounds in Figures 7-8 for time-limited runs) controls only the discretization/Monte Carlo gap between grid-evaluated risk and the true supremum; (ii) for results where the algorithm completes J(epsilon) iterations, the optimization error is controlled by Theorem 1, so the overall certification is valid; and (iii) for time-limited runs where J(epsilon) is not reached, the optimization error is uncontrolled by our theory, and thus the reported upper bound should not be interpreted as a formal epsilon-minimax guarantee. We will also adjust the labeling in the figures and surrounding text to make this distinction clear to readers. revision: yes

Circularity Check

0 steps flagged

No circularity found: the derivation chain is self-contained against external benchmarks

full rationale

I traced the paper's main derivation chain: Lemma 1 (Lipschitz risk) uses an external total variation bound (Devroye et al. 2018); Lemma 2 (discretization error) follows from Lemma 1 and the minimax theorem (Johnstone 2019, Blackwell and Girshick 1954); Theorem 1 (approximate least-favorable distribution) applies the mirror descent regret bound from Bubeck (2015, inequality 4.10) with a Hoeffding-type concentration argument (Lemma 3), both of which are standard external results applied correctly. The supergradient g(π) = (R(d_π, θ_1),...,R(d_π, θ_I)) is computed from the model definition (Bayes risk under N(θ_i,1)), not fitted to the target output. The parameter choices η(ε)=ε/(2M²), J(ε)=⌈2M²ln(I)/(ε/2)²⌉, I(ε)=⌈1+(3m³+4m²)/(ε/2)⌉ are derived algebraically to cancel the deterministic and stochastic error terms to ε/2 each, which is a standard optimization-theory calculation, not a fit renamed as a prediction. Theorems 2–4 (Wasserstein convergence, stochastic minimax sequence, uniform convergence over compacts) follow from Theorem 1 plus the Lipschitz properties of f and d_π established in Lemmas 6–8. The choice ε(m) = (1/5)m²/(1+m²) is motivated by the Ibragimov-Has'minskii constant (Donoho et al. 1990) to ensure ε is below 25% of the minimax risk, not fitted to the output. The paper does cite several works co-authored by Montiel Olea (Aradillas Fernández et al. 2025a,b; Montiel Olea et al. 2026; Blanchet et al. 2023), but these citations are used for context, motivation, or the application section (LP-VAR aggregation), not as load-bearing premises for Theorems 1–4. The minimax theorem and uniqueness of the least-favorable distribution are attributed to Johnstone (2019), an external textbook. The algorithm's output is validated against the closed-form solution of Casella and Strawderman (1981) for m≤1 (Appendix D.4), providing an external benchmark. No step in the central derivation reduces to its inputs by construction, definition, or self-citation chain.

Axiom & Free-Parameter Ledger

2 free parameters · 6 axioms · 0 invented entities

The paper introduces no new mathematical entities, particles, forces, or postulated objects. The 'stochastic minimax sequence' (Definition 1) is a new definition but is a formalization of an existing concept (Ghosh 1964's minimax sequence extended to allow stochasticity), not an invented entity. The algorithm itself is an application of existing mirror descent machinery. The free parameter epsilon has a principled selection rule. The domain assumptions (bivariate normality, bias bound, invariance) are standard in the LP-VAR literature and motivated by Montiel Olea et al. (2026).

free parameters (2)
  • epsilon (approximation error target) = epsilon(m) = (1/5) * m^2 / (1 + m^2)
    The only user-chosen parameter. The paper provides a principled rationale (Section 4.1): set epsilon below 0.25 * v*(m) so the minimax linear estimator is not already epsilon-minimax. All other parameters (step size, iterations, grid size) are determined by epsilon via Theorem 1.
  • M (Lipschitz constant of objective function) = M = 4*m^2
    Derived from the bound |R(d,theta)| <= 4m^2, not fitted to data. Used to set step size and iteration count. The paper notes this is conservative (Appendix D, Figure 15).
axioms (6)
  • standard math The minimax theorem holds for the Bounded Normal Mean problem: v_bar*(m) = v*(m).
    Invoked in Section 2.2 and Appendix A.2. Cited to Proposition 4.19 and Theorem A.5 in Johnstone (2019). Standard result in statistical decision theory.
  • standard math The least-favorable distribution exists and is unique, with finite support.
    Invoked in Theorem 2's proof (Appendix A.4) and footnoted in Section 2.2. Cited to Proposition 4.19 in Johnstone (2019).
  • standard math The minimax theorem holds for the discretized problem (S-games).
    Invoked in Appendix A.2. Cited to Theorem 2.4.2 in Blackwell and Girshick (1954) and Appendix C of Aradillas Fernández et al. (2025b).
  • standard math Mirror descent convergence bound (Bubeck 2015, inequality 4.10) applies to the simplex with negative entropy mirror map.
    Invoked in Appendix A.3, equation (26). Standard result in convex optimization.
  • domain assumption Bivariate normality of (LP estimator, VAR estimator) with known covariance, and a known bound on VAR bias.
    Invoked in Section 5.1, equation (18). Asymptotic normality is motivated by the locally misspecified VAR model of Montiel Olea et al. (2026). The bias bound comes from restricting the fraction of residual variance explained by misspecification to at most M=1%.
  • domain assumption Invariance restriction on estimators: d_h(s+s', t) = d_h(s, t) + s' for all s, s', t.
    Invoked in Section 5.2, equation (20). Follows Armstrong et al. (2025). Restricts to estimators that shift by s' when the LP estimator shifts by s'.

pith-pipeline@v1.1.0-glm · 36306 in / 3910 out tokens · 544304 ms · 2026-07-07T15:27:43.219598+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

61 extracted references · 61 canonical work pages · 6 internal anchors

  1. [1]

    Blanchet, J

    Aradillas Fern \'a ndez, A., J. Blanchet, J. L. Montiel Olea, C. Qiu, J. Stoye, and L. Tan (2025 a ): Approximate least-favorable distributions and nearly optimal tests via stochastic mirror descent, Preprint, arXiv:2511.16925

    work page arXiv 2025

  2. [2]

    --- -.1pt --- -.1pt --- (2025 b ): Epsilon-Minimax Solutions of Statistical Decision Problems, Preprint, arXiv:2509.08107

    work page arXiv 2025

  3. [3]

    Armstrong, T. B., P. Kline, and L. Sun (2025): Adapting to Misspecification, Econometrica, 93, 1981--2005

    work page 2025

  4. [4]

    Barsov, S. S. and V. V. Ulyanov (1987): Estimates of the Proximity of Gaussian Measures, Soviet Mathematics Doklady, 34, 462--466

    work page 1987

  5. [5]

    Margalit, and A

    Ben-Tal, A., T. Margalit, and A. Nemirovski (2001): The ordered subsets mirror descent optimization method with applications to tomography, SIAM Journal on Optimization, 12, 79--108

    work page 2001

  6. [6]

    Bickel, P. J. (1981): Minimax estimation of the mean of a normal distribution when the parameter space is restricted, The Annals of Statistics, 9, 1301--1309

    work page 1981

  7. [7]

    Blackwell, D. and M. A. Girshick (1954): Theory of games and statistical decisions, New York: Wiley

    work page 1954

  8. [8]

    Blanchet, J., Y. Kang, J. L. Montiel Olea, V. A. Nguyen, and X. Zhang (2023): Dropout Training is Distributionally Robust Optimal, Journal of Machine Learning Research, 24, 1--60

    work page 2023

  9. [9]

    Lugosi, and P

    Boucheron, S., G. Lugosi, and P. Massart (2013): Concentration Inequalities: A Nonasymptotic Theory of Independence, Oxford University Press

    work page 2013

  10. [10]

    (2015): Convex Optimization: Algorithms and Complexity, Foundations and Trends in Machine Learning, 8, 231--357

    Bubeck, S. (2015): Convex Optimization: Algorithms and Complexity, Foundations and Trends in Machine Learning, 8, 231--357

    work page 2015

  11. [11]

    Casella, G. and W. E. Strawderman (1981): Estimating a Bounded Normal Mean, The Annals of Statistics, 9, 870--878

    work page 1981

  12. [12]

    Cattaneo, M. D., J. M. Klusowski, and B. Shigida (2024): On the Implicit Bias of Adam, in Forty-first International Conference on Machine Learning

    work page 2024

  13. [13]

    Cattaneo, M. D. and B. Shigida (2025 a ): How Memory in Optimization Algorithms Implicitly Modifies the Loss, in The Thirty-ninth Annual Conference on Neural Information Processing Systems

    work page 2025

  14. [14]

    --- -.1pt --- -.1pt --- (2025 b ): Modified Loss of Momentum Gradient Descent: Fine-Grained Analysis, Preprint, arXiv:2509.08483

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  15. [15]

    (2000): Econometric applications of maxmin expected utility, Journal of Applied Econometrics, 15, 625--644

    Chamberlain, G. (2000): Econometric applications of maxmin expected utility, Journal of Applied Econometrics, 15, 625--644

    work page 2000

  16. [16]

    Chandrasekaran, V. and M. I. Jordan (2013): Computational and statistical tradeoffs via convex relaxation, Proceedings of the National Academy of Sciences, 110, E1181--E1190

    work page 2013

  17. [17]

    Estimator Averaging of Local Projection and VAR Impulse Responses

    Chen, C., E. Pesavento, and B. Vonn \'a k (2026): Estimator Averaging of Local Projection and VAR Impulse Responses, Preprint, arXiv:2605.05456

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  18. [18]

    (2026): Some existence results for maximin priors in statistical minimax theorems, Working paper, University of Southern California

    Delatte, P. (2026): Some existence results for maximin priors in statistical minimax theorems, Working paper, University of Southern California

    work page 2026

  19. [19]

    The total variation distance between high-dimensional Gaussians with the same mean

    Devroye, L., A. Mehrabian, and T. Reddad (2018): The total variation distance between high-dimensional Gaussians with the same mean, Preprint, arXiv:1810.08693

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  20. [20]

    Donoho, D. L., R. C. Liu, and B. MacGibbon (1990): Minimax Risk Over Hyperrectangles, and Implications, The Annals of Statistics, 18, 1416--1437

    work page 1990

  21. [21]

    Hazan, and Y

    Duchi, J., E. Hazan, and Y. Singer (2011): Adaptive Subgradient Methods for Online Learning and Stochastic Optimization, Journal of Machine Learning Research, 12, 2121--2159

    work page 2011

  22. [22]

    Elliott, G., U. K. M \"u ller, and M. W. Watson (2015): Nearly optimal tests when a nuisance parameter is present under the null hypothesis, Econometrica, 83, 771--811

    work page 2015

  23. [23]

    Feldman, I. and L. D. Brown (1989): The minimax risk for estimating a bounded normal mean, Tech. rep., Cornell University, Statistics Center

    work page 1989

  24. [24]

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

    work page 1967

  25. [25]

    Francis, N., M. T. Owyang, J. E. Roush, and R. DiCecio (2014): A Flexible Finite-Horizon Alternative to Long-Run Restrictions with an Application to Technology Shocks, Review of Economics and Statistics, 96, 638--647

    work page 2014

  26. [26]

    Gertler, M. and P. Karadi (2015): Monetary Policy Surprises, Credit Costs, and Economic Activity, American Economic Journal: Macroeconomics, 7, 44--76

    work page 2015

  27. [27]

    Ghosh, M. N. (1964): Uniform approximation of minimax point estimates, The Annals of Mathematical Statistics, 35, 1031--1047

    work page 1964

  28. [28]

    Jaumard, and B

    Gourdin, \'E ., B. Jaumard, and B. MacGibbon (1994): Global optimization decomposition methods for bounded parameter minimax risk evaluation, SIAM Journal on Scientific Computing, 15, 16--35

    work page 1994

  29. [29]

    Guggenberger, P. and J. Huang (2025): On the numerical approximation of minimax regret rules via fictitious play, Preprint, arXiv:2503.10932

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  30. [30]

    Hausman, J. A. (1978): Specification tests in econometrics, Econometrica, 46, 1251--1271

    work page 1978

  31. [31]

    Ibragimov, I. A. and R. Z. Has'minskii (1985): On nonparametric estimation of the value of a linear functional in Gaussian white noise, Theory of Probability & Its Applications, 29, 18--32

    work page 1985

  32. [32]

    Johnstone, I. M. (2019): Gaussian Estimation: Sequence and Wavelet Models, Draft version, September 16, 2019

    work page 2019

  33. [33]

    Johnstone, I. M. and K. B. MacGibbon (1992): Minimax estimation of a constrained Poisson vector, The Annals of Statistics, 20, 807--831

    work page 1992

  34. [34]

    (2005): Estimation and Inference of Impulse Responses by Local Projections, American Economic Review, 95, 161--182

    Jord \`a , \`O . (2005): Estimation and Inference of Impulse Responses by Local Projections, American Economic Review, 95, 161--182

    work page 2005

  35. [35]

    Jord \`a , \`O . and A. M. Taylor (2025): Local projections, Journal of Economic Literature, 63, 59--110

    work page 2025

  36. [36]

    Rigollet, and A

    Juditsky, A., P. Rigollet, and A. B. Tsybakov (2008): Learning by mirror averaging, The Annals of Statistics, 36, 2183--2206

    work page 2008

  37. [37]

    Kempthorne, P. J. (1987): Numerical specification of discrete least favorable prior distributions, SIAM Journal on Scientific and Statistical Computing, 8, 171--184

    work page 1987

  38. [38]

    Kent, C., J. Li, J. Blanchet, and P. W. Glynn (2021): Modified Frank Wolfe in probability space, in Advances in Neural Information Processing Systems (NeurIPS), vol. 34, 14448--14462

    work page 2021

  39. [39]

    Balandat, C

    Krichene, W., M. Balandat, C. Tomlin, and A. Bayen (2015): The hedge algorithm on a continuum, in International Conference on Machine Learning, PMLR, vol. 37, 824--832

    work page 2015

  40. [40]

    Lei, L. and M. I. Jordan (2020): On the Adaptivity of Stochastic Gradient-Based Optimization, SIAM Journal on Optimization, 30, 1473--1500

    work page 2020

  41. [41]

    Levit, B. Y. (1981): On asymptotic minimax estimates of the second order, Theory of Probability & Its Applications, 25, 552--568

    work page 1981

  42. [42]

    Plagborg-M ller, and C

    Li, D., M. Plagborg-M ller, and C. K. Wolf (2024): Local projections vs. VARs: Lessons from thousands of DGPs, Journal of Econometrics, 244, 105722

    work page 2024

  43. [43]

    Lin, Z., P. J. Bickel, and P. Ding (2026): Introducing the b-value: combining unbiased and biased estimators from a sensitivity analysis perspective, Preprint, arXiv:2602.16310

    work page arXiv 2026

  44. [44]

    Manski, C. F. (2021): Econometrics for decision making: Building foundations sketched by Haavelmo and Wald, Econometrica, 89, 2827--2853

    work page 2021

  45. [45]

    Montiel Olea, J. L. and M. Plagborg-M ller (2021): Local Projection Inference is Simpler and More Robust Than You Think, Econometrica, 89, 1789--1823

    work page 2021

  46. [46]

    Montiel Olea, J. L., M. Plagborg-M ller, E. Qian, and C. K. Wolf (2025): Local Projections or VARs? A Primer for Macroeconomists, NBER Macroeconomics Annual, 40, 111--152

    work page 2025

  47. [47]

    --- -.1pt --- -.1pt --- (2026): Double Robustness of Local Projections and Some Unpleasant VAR ithmetic, Econometrica, forthcoming

    work page 2026

  48. [48]

    Bilodeau, N

    Negrea, J., B. Bilodeau, N. Campolongo, F. Orabona, and D. Roy (2021): Minimax optimal quantile and semi-adversarial regret via root-logarithmic regularizers, in Advances in Neural Information Processing Systems (NeurIPS), vol. 34, 26237--26249

    work page 2021

  49. [49]

    Juditsky, G

    Nemirovski, A., A. Juditsky, G. Lan, and A. Shapiro (2009): Robust stochastic approximation approach to stochastic programming, SIAM Journal on Optimization, 19, 1574--1609

    work page 2009

  50. [50]

    Nemirovski, A. and D. Yudin (1983): Problem Complexity and Method Efficiency in Optimization, New York: Wiley

    work page 1983

  51. [51]

    Nemtyrev, A. and O. Boldea (2026): Targeted Local Projections, Preprint, arXiv:2603.00248

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  52. [52]

    Plagborg-M ller, M. and C. K. Wolf (2021): Local Projections and VARs Estimate the Same Impulse Responses, Econometrica, 89, 955--980

    work page 2021

  53. [53]

    Ramey, V. A. (2011): Identifying Government Spending Shocks: It's all in the Timing, Quarterly Journal of Economics, 126, 1--50

    work page 2011

  54. [54]

    --- -.1pt --- -.1pt --- (2016): Macroeconomic Shocks and Their Propagation, in Handbook of Macroeconomics, ed. by J. B. Taylor and H. Uhlig, Elsevier, vol. 2, chap. 2, 71--162

    work page 2016

  55. [55]

    Romer, C. D. and D. H. Romer (2010): The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks, American Economic Review, 100, 763--801

    work page 2010

  56. [56]

    Sims, C. A. (1980): Macroeconomics and Reality, Econometrica, 48, 1--48

    work page 1980

  57. [57]

    Stock, J. H. and M. W. Watson (2018): Identification and estimation of dynamic causal effects in macroeconomics using external instruments, The Economic Journal, 128, 917--948

    work page 2018

  58. [58]

    (2021): Topics in optimal transportation, Graduate Studies in Mathematics, American Mathematical Society

    Villani, C. (2021): Topics in optimal transportation, Graduate Studies in Mathematics, American Mathematical Society

    work page 2021

  59. [59]

    Williamson, D. P. and D. B. Shmoys (2011): The design of approximation algorithms, Cambridge University Press

    work page 2011

  60. [60]

    The Illusion of Learning from Observational Data: An Empirical Bayes Perspective

    Wu, B., S. Salazar, D. P. Green, and D. M. Blei (2026): The Illusion of Learning from Observational Data: An Empirical Bayes Perspective, Preprint, arXiv:2604.08853

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  61. [61]

    (2026): Local Projection-Based Inference under General Conditions , The Review of Economic Studies, rdag034

    Xu, K.-L. (2026): Local Projection-Based Inference under General Conditions , The Review of Economic Studies, rdag034

    work page 2026