Pith. sign in

REVIEW 6 major objections 4 minor 39 references

Distributionally Robust Recovery of Omitted Factors from Forecast Residuals with Application to Interest Rate Risk Management

T0 review · 6 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read A distributionally robust ensemble of factor and random-forest models delivers the best short-horizon U.S. Treasury yield forecasts, while random forests alone dominate at longer horizons.

desk verdict Solid empirical yield-forecasting horse race, but the arXiv abstract doesn't match the full text and the long-horizon claim is not backed by any combination tables. read the letter →

arxiv 2601.04608 v3 pith:KMUHTDVZ submitted 2026-01-08 q-fin.MF q-fin.CPstat.ML

Jinjun Liu , Ming-Yen Cheng This is my paper · ORCID
classification q-fin.MFq-fin.CPstat.ML MSC 91G3062M2062M10
keywords distributionallyrobustoptimizationforecastcombinationTreasuryyieldcurverandomforestsdynamicNelson-Siegelmodelexpectedshortfallensembleforecastingtermstructure
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 recasts yield-curve forecasting as a decision under distributional uncertainty: instead of picking the single best model, the forecaster combines a parametric factor model and a random-forest model with weights that minimize worst-case loss. The central claim is that adaptive, distributionally robust combinations—weights penalizing expected shortfall and using ridge-regularized covariance—outperform individual models at one-to-three-month horizons, while the random forest alone remains the best choice at six-to-twelve-month horizons. The evidence is monthly U.S. Treasury zero-coupon yields from 2006 to 2025, plus a smaller international extension, and the patterns are strongest when the model pool mixes factor and machine-learning forecasts. The paper also claims that these robust weights shift away from the factor model during stress episodes, which stabilizes forecast errors in crises such as the COVID-19 shock and the post-2022 tightening cycle.

What carries the argument

The machinery is the distributionally robust combination layer. Three schemes implement it: FC-DRO-ES (exponential reweighting by expected shortfall at the 10% tail), FC-DRO-MIX (a convex blend of mean squared error and expected shortfall of squared errors), and FC-DRMV (minimum-variance weights with ridge-regularized covariance, tau=0.05). These sit on top of the two base forecast families—the rolling FADNS model, which iterates a VAR(1) on Nelson-Siegel factors augmented with up to ten principal components from 111 indicators, and random forests with rolling direct multi-step regression. The combination weights are re-estimated each month on a 24-month error window, and the DRO variants sh

What would settle it

Compute one-month-ahead RMSFE for the hybrid FADNS+RF pool with DRO hyperparameters selected on a rolling validation window (or with first-release data) and compare against equal-weight averaging and rank-based weighting; if the DRO variants no longer beat these simple schemes across most maturities, the central robustness claim is not supported.

Watch

Extended reading notes

Core claim

The paper proposes a three-part ensemble for forecasting U.S. Treasury zero-coupon yields: a rolling-window factor-augmented dynamic Nelson-Siegel (FADNS) model as parametric baseline, random forests as nonlinear high-dimensional learners, and a distributionally robust combination layer that sets weights from expected shortfall, a hybrid squared-error-plus-tail loss, or ridge-regularized covariance. The central claim is that these adaptive robust combinations beat every individual model at one-to-three-month horizons and that random forests alone dominate at six-to-twelve-month horizons, with the DRO weights visibly reallocating from FADNS to random forests during stress episodes such as the

Load-bearing premise

The empirical case for the DRO combinations rests on point-wise RMSFE comparisons with fixed hyperparameters (eta=5.0, lambda=0.5, tau=0.05) and final-release, linearly interpolated macro data, so if those settings were tuned on the evaluation sample or real-time data arrive differently, the short-horizon advantage could disappear.

Editorial extensions

If this is right

  • For forecasting horizons of one to three months, combining heterogeneous factor and machine-learning forecasts with tail-robust weights outperforms any single model in the pool; users with short decision horizons should prefer the DRO combinations.
  • For six- to twelve-month horizons, the random forest alone is the better choice; recursive factor-model errors accumulate too fast for factor-based combinations to help.
  • Combination rules matter only when the underlying models are heterogeneous; within a pool of random forests alone, all combination schemes perform nearly alike.
  • During extreme market regimes, DRO weights reallocate quickly toward the more robust model family, which smooths forecast-error paths compared with adaptive or classic weighting.
  • The same random-forest specification transfers to 10-year benchmark yields in Canada, China, Germany, Japan, Malaysia, the UK, and the US with stable RMSFE, indicating the approach generalizes beyond zero-coupon Treasuries.

Reading between the lines

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

  • The appended abstract describes a separate framework—recovering omitted factors from forecast residuals via a 'covariance forcing' discovery statistic—that the full text never implements; a reader should not infer that the factor-recovery claims are tested here.
  • The short-horizon gains of the DRO combinations are established only by point-wise RMSFE tables with fixed hyperparameters (eta=5.0, lambda=0.5, tau=0.05); a natural test is to tune these hyperparameters on a rolling validation split and see whether the gains persist out-of-sample.
  • Because the macro panel is final-release and interpolated, the real-time value of the forecasts is untested; re-running with vintage data would clarify whether the robustness survives information delays.
  • The long-horizon dominance of random forests is asserted within the paper's own model families; comparing against a naive random-walk or AR benchmark would place the claim in a wider context.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

6 major / 4 minor

Summary. The manuscript develops a distributionally robust ensemble forecasting framework for U.S. Treasury yields, combining rolling-window Factor-Augmented Dynamic Nelson–Siegel (FADNS) models and high-dimensional Random Forest (RF) models through a battery of forecast combination schemes, including expected-shortfall-based reweighting (FC-DRO-ES, FC-DRO-MIX) and ridge-regularized covariance weighting (FC-DRMV). Using monthly zero-coupon yields from 2006 to 2025 and 111 macroeconomic indicators, the paper reports out-of-sample RMSFE for horizons h=1,3,6,9,12 for individual models, but only for h=1 for the combination methods. The central claims are that adaptive distributionally robust combinations outperform individual models at short horizons and that RF forecasts dominate at medium and longer horizons. Additional sections cover SHAP interpretation, structural break detection, and an international extension of the RF model. The manuscript, however, is internally inconsistent: the arXiv metadata abstract describes an omitted-factor recovery method with covariance forcing, a topic that does not appear anywhere in the full text, whose own abstract and title concern forecast combination.

Significance. If the empirical claims were adequately supported, the paper would offer a practically useful robust combination procedure for term-structure forecasting, and the breadth of the comparison—14 combination rules, 15 maturities, 5 horizons, 111 predictors—would be a useful reference point. The rolling-window out-of-sample design and the explicit use of lagged predictors are genuine strengths, as is the multi-country robustness exercise for RF. However, the paper provides no formal DRO derivation, no statistical tests for forecast differences, no sensitivity analysis for key hyperparameters, and no combination results for h>1; the central empirical claims are therefore not currently established. The manuscript also does not supply code or machine-checked proofs, so the empirical findings rest entirely on point estimates in tables.

major comments (6)
  1. [§3.1.3, Tables 9–10; §5] The conclusion that "RF forecasts dominate at medium and longer horizons" is not supported by any reported results for the forecast combination methods at h=3, 6, 9, or 12. Tables 9 and 10 are both explicitly labeled "horizon h=1 month," and the E-Companion contains no RMSFE tables for the combination schemes at longer horizons. The claim can only be checked for individual FADNS vs. RF; it does not establish that the combinations—which at h=1 often outperform RF—fail to do so at longer horizons. Please report combination results at all horizons, or explicitly restrict the conclusions.
  2. [Tables 9–10 vs. Table 1] The claimed superiority of the DRO combinations is based on point estimates without measures of uncertainty. For example, FC-DRO-MIX at the 3M maturity has RMSFE 21.18 bps in Table 9, which lies inside the RF seed range [17.20, 32.40] from Table 1; similarly, many point differences across the 14 combination rules are small relative to plausible sampling variation. A Diebold–Mariano test, a block bootstrap, or another procedure accounting for time-series dependence and multiple comparisons across 14 methods and 15 maturities is needed before any claim of "systematically improved performance" can be accepted.
  3. [§2.4.4, §2.4.3, §2.4] Key hyperparameters—η=5.0, λ=0.5, τ=0.05, W=24, L=20, α=0.10, φ_n=0.02—are fixed without sensitivity analysis and without any validation split. If these values were chosen or adjusted based on the h=1 evaluation sample, the out-of-sample interpretation of the results collapses. The manuscript needs either a formal validation-based selection procedure or a sensitivity analysis demonstrating that the qualitative conclusions are robust over reasonable hyperparameter ranges.
  4. [§3.1.2, Table 8] Table 8 reports the "best PCA dimension" for each maturity and horizon. If this selection was made by comparing full-sample RMSFE across the ten FADNS specifications, then the reported FADNS performance is in-sample-selected and the hybrid pool of "10 FADNS models" used in combination may inherit this selection. The paper must clarify whether the best PCA dimension is chosen recursively using only information available at each forecast origin, or acknowledge that the FADNS results are optimistic. This issue is load-bearing because FADNS is one of the two families in the combination pool.
  5. [§2.4.4, §2.4, Abstract] The methods labeled "Distributionally Robust" are not derived from a well-posed DRO problem. No ambiguity set is defined for the combination weights, no worst-case expectation is minimized, and no theorem connects the ES-based exponential reweighting or ridge-regularized covariance to a Delage–Ye-type moment ambiguity set. As written, FC-DRO-ES and FC-DRO-MIX are heuristic reweighting rules, and FC-DRMV is a ridge-regularized minimum-variance combination. Either provide a formal equivalence (e.g., show that the ES penalty is the dual of an appropriate moment constraint), or revise the terminology and claims so that the contribution is not overstated.
  6. [Title, Abstract, Full Text] The manuscript is internally inconsistent at the level of its identity. The arXiv metadata abstract describes a paper on "Distributionally Robust Recovery of Omitted Factors from Forecast Residuals," with covariance forcing, factor naming, and block-permutation certification; the full text is a paper on forecasting the U.S. Treasury yield curve with FADNS, RF, and forecast combinations, and none of the recovery content appears. This is not a minor typo: a reader cannot tell which paper is being submitted. The authors must reconcile the abstract/title with the actual content.
minor comments (4)
  1. [Algorithm OA.4] Step (8) is labeled "FC-JMA" but the corresponding method in §2.4.2 is called "FC-LAD"; step (14) also refers to "FC-JMA". The labels should be harmonized.
  2. [§4.2] The international extension applies only to the Random Forest model. The abstract and conclusion claim that the "framework" generalizes, but no forecast combination or DRO method is tested globally. Please either add such results or soften the generalization claim.
  3. [§3.2] The assertion that DRO combinations show "markedly smoother error dynamics" is qualitative and not quantified. Report, for example, the standard deviation or interquartile range of forecast errors across methods, or present a formal comparison.
  4. [§2.1.2] The use of final-release data and linear interpolation of quarterly variables is acknowledged as a limitation, but this should be revisited in the main conclusions: it weakens the claim that the forecasts reflect real-time information availability.

Circularity Check

1 steps flagged · score 4.0 of 10

Full-text DRO forecasting derivation is self-contained; the prefixed abstract's neutralization claim reduces to the definition of the recovered factor.

  1. self definitional [Prefixed Abstract (immediately before Full Text; not present in Full Text body)]
    "the discovery statistic is the covariance forcing, the component of the decision's residual risk transverse to its exposure. ... A neutralization test completes the evidence: removing the recovered factor from the deployed duration position leaves volatility essentially unchanged and worsens the tail, so the factor is a material systematic risk the position bears."

    The factor is defined as the component of residual risk transverse (orthogonal) to the position's exposure. Removing an orthogonal component cannot change the position's volatility, so 'leaves volatility essentially unchanged' is true by construction, not by empirical test. Presenting this as evidence that the factor is material makes the evidence equivalent to the definition. The Full Text body contains no such recovery/neutralization analysis; if this abstract is part of the manuscript, this is a self-definitional step.

full rationale

The yield-curve forecasting body is otherwise self-contained: FADNS and RF forecasts are generated recursively from rolling windows, and the DRO combination weights (FC-DRO-ES, FC-MIX, FC-DRMV) are computed only from past forecast errors (W=24) and then evaluated on out-of-sample data, so no target quantity is used to define the predictors. The Jiang et al. (2025) self-citation for the LAD combination is minor and not load-bearing. The unsupported longer-horizon claim and the absence of h>1 combination tables are evidence gaps, not circularity. The only definitional reduction found is the neutralization statement in the prefixed abstract, which is why the score is above 2 but far below a wholly circular derivation.

Assumptions & free parameters 9 free parameters · 6 assumptions · 0 invented entities

The central empirical claims rest on a large set of hand-chosen hyperparameters (η, λ, τ, φ, α, window lengths) and on domain assumptions about data timeliness and rolling-window stability. No new theoretical entities are introduced.

free parameters (9)
  • η (robustness temperature in FC-DRO-ES and FC-MIX) = 5.0
    Fixed across maturities and horizons; controls the sharpness of exponential reweighting; no sensitivity analysis reported.
  • λ (mix weight in FC-MIX) = 0.5
    Balances MSE and expected shortfall; fixed by hand.
  • τ (ridge parameter in FC-DRMV) = 0.05
    Stabilizes covariance inversion; fixed without validation.
  • φ_n (LAD penalty) = 0.02
    Penalty for weight sparsity in FC-LAD; fixed.
  • α (expected shortfall level) = 0.10
    Tail level for ES and squared-error ES.
  • L (AFTER lookback) = 20
    Lookback length for AFTER variance estimation.
  • W (combination window length) = 24 months
    Rolling window for computing forecast-error moments; a key modeling choice.
  • w (training window length) = 60 months
    Rolling training window for DNS/FADNS/RF models.
  • best PCA dimension k = per maturity/horizon (Table 8)
    Selected post hoc as the dimension with lowest out-of-sample RMSFE; a model-selection free parameter.
assumptions (6)
  • domain assumption DNS factors follow a Gaussian VAR(1): η_t ~ N(0, Σ_η) (Eq 2.2b)
    Used to generate multi-horizon forecasts by iterating the VAR; misspecification would affect FADNS/DNS forecasts.
  • domain assumption Macroeconomic predictors, after ADF-based differencing and standardization, have stable second moments within each 60-month rolling window
    Rolling PCA factors and VAR estimates assume within-window covariance stationarity despite known structural breaks.
  • domain assumption Final-release data adequately represents the real-time information set
    Acknowledged in Section 2.1.2; real-time forecasters would have first-release data, changing predictor values and forecast errors.
  • domain assumption Quarterly variables linearly interpolated to monthly frequency preserve cyclical information
    Acknowledged as possibly attenuating short-run fluctuations (Section 2.1.2).
  • domain assumption A 24-month rolling window yields reliable expected-shortfall and covariance estimates
    Window length W=24 is short; during regime shifts, moment estimates may be unrepresentative.
  • ad hoc to paper Penalizing expected shortfall and ridge-regularizing covariance implements a distributionally robust worst-case criterion
    The paper does not construct a Delage–Ye ambiguity set or solve a min-max problem; the connection to DRO is asserted, not derived.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distributionally Robust Recovery of Omitted Factors from Forecast Residuals with Application to Interest Rate Risk Management." pith.science (2026). https://pith.science/paper/KMUHTDVZ

@misc{pith2026260104608,
  author       = {Pith},
  title        = {Pith review of: Distributionally Robust Recovery of Omitted Factors from Forecast Residuals with Application to Interest Rate Risk Management},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KMUHTDVZ}},
  note         = {Machine review of arXiv:2601.04608}
}
read the original abstract

A forecasting model compresses its predictors into an estimate of a conditional mean, and the systematic structure that estimate omits survives in the second moment of its forecast errors. Accuracy comparisons do not measure this structure, and variance-based extraction does not recover the part of it that a given decision bears. In this paper, we propose a distributionally robust framework that recovers the omitted structure from the residuals of a fixed forecaster: a decision is made robust over a two-layer moment ambiguity set on the standardized residual cross-section, and the discovery statistic is the covariance forcing, the component of the decision's residual risk transverse to its exposure. We demonstrate that the forcing is invariant to shrinkage and to isotropic inflation of the covariance, so the recovered direction is a property of the residuals rather than of the regularization, the sense in which the recovery is ground truth. This is confirmed on monthly U.S. Treasury zero-coupon yields from 2006 to 2025, where the recovered factor is named by factor-adjusted robust selection against a panel of 111 macroeconomic, Treasury supply-and-demand, and financial indicators. From the residuals of the linear factor-augmented dynamic Nelson-Siegel benchmark the factor names as a leading business-cycle factor, anchored on the Conference Board leading index and certified by a block-permutation test; from those of the more accurate nonlinear random forest benchmark the same procedure selects the same real-activity family without certification. A neutralization test completes the evidence: removing the recovered factor from the deployed duration position leaves volatility essentially unchanged and worsens the tail, so the factor is a material systematic risk the position bears.

Figures

Figures reproduced from arXiv: 2601.04608 by the authors.

Figure 1
Figure 1. Comparing U.S. benchmark Treasury yield forecasts with additional Treasury supply [PITH_FULL_IMAGE:figures/full_fig_p024_1.png] view at source ↗
Figure 2
Figure 2. Maturity-averaged global SHAP values for the Random Forest model across forecast [PITH_FULL_IMAGE:figures/full_fig_p030_2.png] view at source ↗
Figure 3
Figure 3. One-month-ahead forecast error dynamics of hybrid RF–FADNS forecast combinations [PITH_FULL_IMAGE:figures/full_fig_p030_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Weight dynamics under distributionally robust mean–variance (DRMV) forecast combi [PITH_FULL_IMAGE:figures/full_fig_p031_4.png]
Figure 5
Figure 5. Figure 5: Weight dynamics under distributionally robust expected shortfall (DRO–ES) forecast [PITH_FULL_IMAGE:figures/full_fig_p031_5.png]
Figure 6
Figure 6. Figure 6: Weight dynamics under hybrid distributionally robust (DRO–MIX) forecast combination. [PITH_FULL_IMAGE:figures/full_fig_p032_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 2 linked inside Pith

  1. [1]

    M., and S.-I

    Lundberg, S. M., and S.-I. Lee. 2017. A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 4765--4774

    2017

  2. [2]

    Timmermann

    Aiolfi, M., and A. Timmermann. 2006. Persistence in forecasting performance and conditional combination strategies. Journal of Econometrics, 135 (1--2), 31--53

    2006

  3. [3]

    Albuquerque, P. H. M., Y. Peng, and J. P. F. Silva. 2022. Making the whole greater than the sum of its parts: A literature review of ensemble methods for financial time series forecasting. Journal of Forecasting, 41 (8), 1701--1724

    2022

  4. [4]

    Piazzesi

    Ang, A., and M. Piazzesi. 2003. A no-arbitrage vector autoregression of term structure dynamics with macroeconomic and latent variables. Journal of Monetary Economics, 50 (4), 745--787

    2003

  5. [5]

    M., and C

    Bates, J. M., and C. W. J. Granger. 1969. The combination of forecasts. Operations Research Quarterly, 20 (4), 451--468

    1969

  6. [6]

    Botchkarev, A. 2019. Performance metrics (error measures) in machine learning regression, forecasting and prognostics: Properties and typology. Interdisciplinary Journal of Information, Knowledge, and Management, 14, 45--79

    2019

  7. [7]

    Breiman, L. 2001. Random forests. Machine Learning, 45 (1), 5--32

    2001

  8. [8]

    Clemente, and T

    Caldeira, J., G. Clemente, and T. Rebbeck. 2016. Forecast combination in the term structure of interest rates: Are machine learning and bagging able to improve. Journal of Forecasting, 35 (4), 323--340

    2016

Show all 39 references
  1. [9]

    Vossler, Y

    Chi, C.-M., P. Vossler, Y. Fan, and J. Lv. 2022. Asymptotic properties of high-dimensional random forests. arXiv preprint arXiv:2004.13953

    2022 arXiv

  2. [10]

    A., and W

    Dickey, D. A., and W. A. Fuller. 1979. Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association, 74 (366), 427--431

    1979

  3. [11]

    X., and C

    Diebold, F. X., and C. Li. 2006. Forecasting the term structure of government bond yields. Journal of Econometrics, 130 (2), 337--364

    2006

  4. [12]

    Fernandes, C., and F. Vieira. 2019. A dynamic Nelson--Siegel model with forward-looking macroeconomic factors for the yield curve in the U.S. Journal of Economic Dynamics and Control, 106, 103720

    2019

  5. [13]

    Granger, C. W. J., and Y. Jeon. 2004. Thick modeling. Economic Modelling, 21 (2), 323--343

    2004

  6. [14]

    Healy, C., and C. Jia. 2023. Monetary policy since the onset of the COVID-19 pandemic: A path-dependent interpretation. Federal Reserve Bank of Cleveland Economic Commentary, 2023-12

    2023

  7. [15]

    R., and A

    Nelson, C. R., and A. F. Siegel. 1987. Parsimonious modeling of yield curves. Journal of Business, 60 (4), 473--489

    1987

  8. [16]

    E., and D

    Said, S. E., and D. A. Dickey. 1984. Testing for unit roots in autoregressive-moving average models of unknown order. Biometrika, 71 (3), 599--607

    1984

  9. [17]

    Scott, D. W. 1992. Multivariate Density Estimation: Theory, Practice, and Visualization. John Wiley & Sons, New York

    1992

  10. [18]

    H., and M

    Stock, J. H., and M. W. Watson. 2002. Forecasting using principal components from a large number of predictors. Journal of the American Statistical Association, 97 (460), 1167--1179

    2002

  11. [19]

    R., and W

    Swanson, N. R., and W. Xiong. 2018. Big data analytics in economics: What have we learned so far, and where should we go from here? Canadian Journal of Economics, 51 (3), 695--746

    2018

  12. [20]

    Wang, X., R. J. Hyndman, F. Li, and Y. Kang. 2023. Forecast combinations: An over 50-year review. International Journal of Forecasting, 39, 1518--1547

    2023

  13. [21]

    Yang, Y. 2004. Combining forecasting procedures: Some theoretical results. Econometric Theory, 20 (1), 176--222

    2004

  14. [22]

    Zou, H., and Y. Yang. 2004. Combining time series models for forecasting. International Journal of Forecasting, 20 (1), 69--84

    2004

  15. [23]

    Brown, R. L., J. Durbin, and J. M. Evans. 1975. Techniques for testing the constancy of regression relationships over time. Journal of the Royal Statistical Society: Series B (Methodological), 37 (2), 149--163

    1975

  16. [24]

    Oudre, and N

    Truong, C., L. Oudre, and N. Vayatis. 2020. Selective review of offline change point detection methods. Signal Processing, 167, 107299

    2020

  17. [25]

    Freund, Y., and R. E. Schapire. 1997. A decision-theoretic generalization of on-line learning and an application to boosting. Journal of Computer and System Sciences, 55 (1), 119--139

    1997

  18. [26]

    Schapire, R. E. 1990. The strength of weak learnability. Machine Learning, 5 (2), 197--227

    1990

  19. [27]

    Freund, Y., and R. E. Schapire. 1996. Experiments with a new boosting algorithm. Proceedings of the Thirteenth International Conference on Machine Learning, 148--156

    1996

  20. [28]

    Freund, Y., and R. E. Schapire. 1999. Large margin classification using the perceptron algorithm. Machine Learning, 37, 277--296

    1999

  21. [29]

    Bai, J., and S. Ng. 2006. Evaluating latent and observed factors in macroeconomics and finance. Journal of Econometrics, 131 (1--2), 507--537

    2006

  22. [30]

    Timmermann, A. 2006. Forecast combinations. Handbook of Economic Forecasting, 1, 135--196

    2006

  23. [31]

    Granger, C. W. J., and R. Ramanathan. 1984. Improved methods of combining forecasts. Journal of Forecasting, 3 (2), 197--204

    1984

  24. [32]

    Wolpert, D. H. 1992. Stacked generalization. Neural Networks, 5 (2), 241--259

    1992

  25. [33]

    Breiman, L. 1996. Stacked regressions. Machine Learning, 24 (1), 49--64

    1996

  26. [34]

    Jiang, X., Y. Lv, Q. Li, and M.-Y. Cheng. 2025. Robust model averaging prediction of longitudinal response with ultrahigh-dimensional covariates. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 87 (2), 337--361

    2025

  27. [35]

    Jin, and Z

    Cartea, \'A., L. Jin, and Z. Shi. 2025. The limited virtue of complexity in a noisy world. University of Oxford and Imperial College London Working Paper

    2025

  28. [36]

    Minorics, and P

    Janzing, D., L. Minorics, and P. Bl \"o baum. 2020. Feature relevance quantification in explainable AI: A causal problem. Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics, 2907--2916

    2020

  29. [37]

    Delage, E., and Y. Ye. 2010. Distributionally robust optimization under moment uncertainty with application to data-driven problems. Operations Research, 58 (3), 595--612

    2010

  30. [38]

    Nguyen, V. A., F. Zhang, J. Blanchet, E. Delage, and Y. Ye. 2020. Distributionally robust local non-parametric conditional estimation. arXiv preprint arXiv:2010.05373

    2020 arXiv

  31. [39]

    Nguyen, V. A., F. Zhang, S. Wang, J. Blanchet, E. Delage, and Y. Ye. 2024. Robustifying conditional portfolio decisions via optimal transport. Operations Research

    2024

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.