pith. sign in

arxiv: 2606.08819 · v1 · pith:4UHPPKIVnew · submitted 2026-06-07 · 📊 stat.ME

Model Selection for SLOPE Models: A Bayesian Perspective

Fabio Feser , Marina Evangelou This is my paper

Pith reviewed 2026-06-27 17:40 UTC · model grok-4.3

classification 📊 stat.ME
keywords SLOPEBayesian model selectionFDR controlspike-and-slab priorvariable selectiongroup selectionhigh-dimensional regression
0
0 comments X

The pith

Bayesian spike-and-slab priors restore FDR control for SLOPE models under general conditions.

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

The paper develops Bayesian Group SLOPE (BGSLOPE) and Bayesian Sparse-group SLOPE (BSGS) by embedding the sorted L1 penalty inside a spike-and-slab prior. It adds the Two-step Orthogonal (TSO) transformation that converts a general regression design into an orthogonal one so that SLOPE's known FDR guarantees apply again. Synthetic and real-data experiments compare these Bayesian selectors against cross-validation and other standard rules, showing that they keep the false discovery rate at the target level, recover more true signals, and produce lower prediction error. A reader would care because most practical data sets violate the orthogonal assumption that makes ordinary SLOPE work, forcing a choice between FDR control and predictive performance.

Core claim

By embedding group-based SLOPE models into a spike-and-slab framework and applying the TSO transformation to recover orthogonality, the proposed Bayesian models achieve FDR control under general conditions while outperforming competing model selection strategies in power and predictive performance.

What carries the argument

The spike-and-slab prior placed around the SLOPE penalty, together with the Two-step Orthogonal (TSO) transformation that converts non-orthogonal designs into orthogonal ones.

If this is right

  • The Bayesian models maintain FDR control when predictors are correlated.
  • They achieve higher true-positive rates than cross-validation at the same FDR level.
  • Predictive performance improves on both simulated and real data sets.
  • The approach covers both ordinary group selection and sparse-group selection.

Where Pith is reading between the lines

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

  • The same spike-and-slab construction could be applied to other sorted penalties to regain theoretical FDR properties.
  • Posterior samples from these models supply direct uncertainty measures that could guide choice of the FDR target.
  • The methods may be especially useful in fields where both error-rate control and accurate prediction are required simultaneously.

Load-bearing premise

Embedding the SLOPE penalty into a spike-and-slab prior and applying the TSO transformation preserves the original FDR-control guarantees under the general non-orthogonal conditions encountered in practice.

What would settle it

A simulation or real-data example on non-orthogonal predictors in which the empirical false discovery rate of BGSLOPE or BSGS exceeds the nominal target level would show the central claim is incorrect.

Figures

Figures reproduced from arXiv: 2606.08819 by Fabio Feser, Marina Evangelou.

Figure 1
Figure 1. Figure 1: BGSLOPE dependency graph. θg θv γ δ β cg cv σ 2 y [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Bayesian latent variables and model parameters for BSGS, SGSLOBE, BGSLOPE, and [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: BH and Gaussian SLOPE se￾quences for different values of q for p = 100, n = 500. Future work should more carefully consider the step 1 model, as it determines TSO’s power. The lasso was chosen for its simplicity and well-understood vari￾able selection properties, and it guarantees |Sˆ v| ≤ n [102]. Ideally, this step would minimize false nega￾tives (type II error), though Su et al. [99] notes this is gener… view at source ↗
Figure 6
Figure 6. Figure 6: FDR (bottom plots) and power (top plots) for the best performing model selection ap [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: OOS error (log scale) for all model selection approaches, as a function of the sparsity [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: MSE(β) (log scale) for all model selection approaches, as a function of the sparsity proportion (top row) and signal strength (bottom row), split into the type of model (SLOPE, gSLOPE, SGS). 5.2.2 Impact of data-generating parameters The impact of changing the dimensionality and noise parameters in the data-generating process is assessed in [PITH_FULL_IMAGE:figures/full_fig_p018_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FDR (bottom plots) and power (top plots) for the best performing model selection ap [PITH_FULL_IMAGE:figures/full_fig_p018_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: MAE(σ) as a function of the noise for all model selection approaches that estimate the noise, split into the type of model (SLOPE, gSLOPE, SGS). The high FDR of Bayesian models under large noise highlights both a challenge and an opportunity: they tend to select overly saturated models. This can be mitigated in BSGS by adjusting the priors on θg and θv to favor sparsity. For instance, setting θg, θv ∼ Bet… view at source ↗
Figure 11
Figure 11. Figure 11: Power and FDR of BSGS under the Scheme 1 priors (BSGS) and [PITH_FULL_IMAGE:figures/full_fig_p019_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: FDR (bottom plots) and power (top plots) for the best performing model selection [PITH_FULL_IMAGE:figures/full_fig_p020_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: MSE(β), variable Power, and variable FDR for all SGS model selection approaches as a function of α under the baseline setting ( [PITH_FULL_IMAGE:figures/full_fig_p022_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Bayesian latent variables and model parameters for BSGS- [PITH_FULL_IMAGE:figures/full_fig_p046_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: The log-likelihood for BSGS, SGSLOBE, BGSLOPE, and GSLOBE for the illustrative [PITH_FULL_IMAGE:figures/full_fig_p047_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: MSE( [PITH_FULL_IMAGE:figures/full_fig_p047_16.png] view at source ↗
Figure 17
Figure 17. Figure 17 [PITH_FULL_IMAGE:figures/full_fig_p048_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: MSE(β) and MAE(σ) for BGSLOPE and GSLOBE under different β initialization models (top two rows) and Beta prior choices (bottom two rows), with a small amount of jitter added to allow the differences to be seen. 48 [PITH_FULL_IMAGE:figures/full_fig_p048_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Distribution of learned [PITH_FULL_IMAGE:figures/full_fig_p049_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: The proportion of instances in which the coin flip property did not hold for active variables, [PITH_FULL_IMAGE:figures/full_fig_p054_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: F1 scores for all model selection approaches, shown for all cases considered, split into the type of model (SLOPE, gSLOPE, SGS). 57 [PITH_FULL_IMAGE:figures/full_fig_p057_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: MAE(σ) for all model selection approaches that estimate the noise, shown for all cases considered, split into the type of model (SLOPE, gSLOPE, SGS). 58 [PITH_FULL_IMAGE:figures/full_fig_p058_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: FDR (bottom plots) and power (top plots) for the other model selection approaches, as [PITH_FULL_IMAGE:figures/full_fig_p059_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: FDR (bottom plots) and power (top plots) for the other model selection approaches, [PITH_FULL_IMAGE:figures/full_fig_p059_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: MSE(β) (log scale) for all model selection approaches, as a function of the dimensionality (top row) and noise (bottom row), split into the type of model (SLOPE, gSLOPE, SGS). 250 500 750 1000 Dimensionality 10 0 10 1 10 2 10 3 10 4 OOS error (log) SLOPE 250 500 750 1000 Dimensionality gSLOPE 250 500 750 1000 Dimensionality SGS 0 1 2 3 Noise 10 0 10 1 10 2 10 3 10 4 OOS error (log) 0 1 2 3 Noise 0 1 2 3 N… view at source ↗
Figure 26
Figure 26. Figure 26: OOS error (log scale) for all model selection approaches, as a function of the dimensionality [PITH_FULL_IMAGE:figures/full_fig_p060_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Runtime in seconds (log scale) for all model selection approaches, as a function of the [PITH_FULL_IMAGE:figures/full_fig_p060_27.png] view at source ↗
Figure 28
Figure 28. Figure 28: FDR (bottom plots) and power (top plots) for the other model selection approaches, as [PITH_FULL_IMAGE:figures/full_fig_p061_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: MSE(β) (log scale) for all model selection approaches, as a function of within-group correlation (top row) and across-group correlation (bottom row), split into the type of model (SLOPE, gSLOPE, SGS). 61 [PITH_FULL_IMAGE:figures/full_fig_p061_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: OOS error (log scale) for all model selection approaches, as a function of within-group [PITH_FULL_IMAGE:figures/full_fig_p062_30.png] view at source ↗
Figure 31
Figure 31. Figure 31: Variable and group FDR shown for CV and BSGS for all synthetic cases considered. [PITH_FULL_IMAGE:figures/full_fig_p062_31.png] view at source ↗
Figure 32
Figure 32. Figure 32: F1 score (top row), power (middle row), and FDR (bottom plots) for all model selection approaches as a function of group size for equal groups. D.5 Real data experiments D.5.1 Dataset information The following real datasets are used in this manuscript: • BRCA1: Gene expression data for breast cancer tissue samples [71]. – Response: Gene expression measurements for the BRCA1 gene. – Data matrix: Gene expre… view at source ↗
read the original abstract

Sorted $\ell_1$ Penalized Estimation (SLOPE) models, that perform either variable or group selection, control the false discovery rate (FDR) under orthogonal settings with known noise, but such settings are rare in practice. Under general conditions, cross-validation is the default model selection approach for SLOPE, yet it targets predictive performance rather than FDR control. We address this gap for the SLOPE family of models by proposing new Bayesian approaches, Bayesian Group SLOPE (BGSLOPE) and Bayesian Sparse-group SLOPE (BSGS). BGSLOPE and BSGS embed group-based SLOPE models into a spike-and-slab framework, with BSGS providing a continuous spike-and-slab framework for sparse-group models. We further introduce Two-step Orthogonal (TSO), which transforms a general setting into an orthogonal one to recover SLOPE's FDR control properties. Through extensive synthetic and real data studies comparing all major model selection strategies for SLOPE models, the proposed Bayesian models consistently control FDR, achieve higher power, and outperform competing methods in prediction.

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 / 1 minor

Summary. The paper proposes Bayesian Group SLOPE (BGSLOPE) and Bayesian Sparse-group SLOPE (BSGS) models that embed SLOPE penalties into spike-and-slab priors, along with a Two-step Orthogonal (TSO) transformation to convert general designs to orthogonal ones. It claims that these approaches recover SLOPE's FDR control properties, deliver higher power than cross-validation and other selectors, and improve prediction, as demonstrated across extensive synthetic and real-data experiments comparing all major model-selection strategies for SLOPE.

Significance. If the FDR-control and performance claims are substantiated, the work would fill a practical gap by supplying Bayesian model-selection tools for SLOPE that target FDR rather than pure prediction error. The continuous spike-and-slab formulation for sparse-group models and the TSO preprocessing step are technically interesting contributions.

major comments (2)
  1. [Theoretical development / TSO section] No derivation is supplied showing that the posterior mode (or thresholded posterior inclusion probabilities) obtained after embedding the SLOPE penalty in a spike-and-slab prior and applying TSO inherits the exact FDR bound of orthogonal SLOPE under non-orthogonal designs; the abstract and theoretical sections assert recovery of the FDR guarantee without a supporting argument once the prior and group structure are introduced.
  2. [Simulation studies] The headline performance claims rest on synthetic studies whose design details (correlation regimes, group-overlap patterns, exact FDR estimation procedure, data-exclusion rules, and number of Monte Carlo replications) are not reported, making it impossible to assess whether the reported FDR control and power gains are robust or merely artifacts of the chosen simulation settings.
minor comments (1)
  1. [Model specification] Notation for the spike-and-slab variances and the TSO transformation matrix should be introduced with explicit definitions before first use.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Theoretical development / TSO section] No derivation is supplied showing that the posterior mode (or thresholded posterior inclusion probabilities) obtained after embedding the SLOPE penalty in a spike-and-slab prior and applying TSO inherits the exact FDR bound of orthogonal SLOPE under non-orthogonal designs; the abstract and theoretical sections assert recovery of the FDR guarantee without a supporting argument once the prior and group structure are introduced.

    Authors: We agree that an explicit derivation linking the TSO-transformed posterior mode to the exact orthogonal SLOPE FDR bound was omitted. In the revision we will add a dedicated theoretical subsection that (i) recalls the FDR guarantee for orthogonal SLOPE, (ii) shows that the two-step orthogonal transformation produces an exactly orthogonal design while preserving the group structure, and (iii) verifies that the chosen spike-and-slab prior is calibrated so its posterior mode recovers the SLOPE estimator (hence the FDR bound) even after the group embedding. The argument will be stated for both BGSLOPE and BSGS. revision: yes

  2. Referee: [Simulation studies] The headline performance claims rest on synthetic studies whose design details (correlation regimes, group-overlap patterns, exact FDR estimation procedure, data-exclusion rules, and number of Monte Carlo replications) are not reported, making it impossible to assess whether the reported FDR control and power gains are robust or merely artifacts of the chosen simulation settings.

    Authors: We accept that the simulation protocol was under-specified. The revised manuscript will expand the simulation section to report: correlation regimes (independent, AR(1) with ho ∈ {0.2,0.5,0.8}, equicorrelated blocks); group-overlap patterns (non-overlapping, 10–30 % partial overlap, full overlap); FDR estimation (Monte-Carlo average of the proportion of false discoveries using the standard definition from the SLOPE literature); data-exclusion rules (none beyond centering/scaling); and number of replications (500 per setting). These additions will allow direct assessment of robustness. revision: yes

Circularity Check

0 steps flagged

No circularity detected; FDR and performance claims rest on external empirical validation rather than self-referential definitions or fitted inputs.

full rationale

The paper's core claims concern empirical performance of BGSLOPE/BSGS plus TSO on synthetic and real data, with FDR control asserted via simulation results rather than any derivation that reduces to the input data or prior fits by construction. No equations equate a 'prediction' to a fitted parameter, no self-citation chain bears the load of a uniqueness theorem, and the TSO transformation is presented as a preprocessing step whose properties are checked externally rather than defined circularly. The derivation chain is therefore self-contained against the reported benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract does not enumerate hyperparameters or priors; typical spike-and-slab constructions introduce slab variance and spike probability parameters that would be fitted or chosen, but none are specified here.

pith-pipeline@v0.9.1-grok · 5711 in / 991 out tokens · 16659 ms · 2026-06-27T17:40:05.993558+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

133 extracted references · 10 canonical work pages · 5 internal anchors

  1. [1]

    Stegun.Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables

    Milton Abramowitz and Irene A. Stegun.Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables. Dover, New York, 1972

    1972

  2. [2]

    False Discovery Rate Estimation for Stability Selection: Application to Genome-Wide Association Studies.Statistical Applications in Genetics and Molecular Biology, 10(1), 2011

    Ismaïl Ahmed, Anna-Liisa Hartikainen, Marjo-Riitta Järvelin, and Sylvia Richardson. False Discovery Rate Estimation for Stability Selection: Application to Genome-Wide Association Studies.Statistical Applications in Genetics and Molecular Biology, 10(1), 2011

    2011

  3. [3]

    H. Akaike. A new look at the statistical model identification.IEEE Transactions on Automatic Control, 19(6):716–723, 1974

    1974

  4. [4]

    A survey of cross-validation procedures for model selection

    Sylvain Arlot and Alain Celisse. A survey of cross-validation procedures for model selection. Statistics Surveys, 4:40–79, 2010

    2010

  5. [5]

    George.Spike-and-Slab Meets LASSO: A Review of the Spike-and-Slab LASSO, pages 81–108

    Ray Bai, Veronika Roˇcková, and Edward I. George.Spike-and-Slab Meets LASSO: A Review of the Spike-and-Slab LASSO, pages 81–108. Chapman and Hall/CRC, 2021

    2021

  6. [6]

    Spike-and- Slab Group Lassos for Grouped Regression and Sparse Generalized Additive Models.Journal of the American Statistical Association, 117(537):184–197, 2022

    Ray Bai, Gemma E Moran, Joseph L Antonelli, Yong Chen, and Mary R Boland. Spike-and- Slab Group Lassos for Grouped Regression and Sparse Generalized Additive Models.Journal of the American Statistical Association, 117(537):184–197, 2022

    2022

  7. [7]

    Controlling the false discovery rate via knockoffs

    Rina Foygel Barber and Emmanuel J Candès. Controlling the false discovery rate via knockoffs. The Annals of Statistics, 43(5), 2015

    2015

  8. [8]

    A knockoff filter for high-dimensional selective inference.The Annals of Statistics, 47(5), 2019

    Rina Foygel Barber and Emmanuel J Candès. A knockoff filter for high-dimensional selective inference.The Annals of Statistics, 47(5), 2019

    2019

  9. [9]

    The p-filter: Multilayer False Discovery Rate Control for Grouped Hypotheses.Journal of the Royal Statistical Society Series B: Statistical Methodology, 79(4):1247–1268, 2017

    Rina Foygel Barber and Aaditya Ramdas. The p-filter: Multilayer False Discovery Rate Control for Grouped Hypotheses.Journal of the Royal Statistical Society Series B: Statistical Methodology, 79(4):1247–1268, 2017

    2017

  10. [10]

    Metropolized Knockoff Sampling.Journal of the American Statistical Association, 116(535):1413–1427, 2021

    Stephen Bates, Emmanuel Candès, Lucas Janson, and Wenshuo Wang. Metropolized Knockoff Sampling.Journal of the American Statistical Association, 116(535):1413–1427, 2021

    2021

  11. [11]

    Cross-validation: What does it estimate and how well does it do it?Journal of the American Statistical Association, 119(546): 1434–1445, 2024

    Stephen Bates, Trevor Hastie, and Robert Tibshirani. Cross-validation: What does it estimate and how well does it do it?Journal of the American Statistical Association, 119(546): 1434–1445, 2024

    2024

  12. [12]

    A fast iterative shrinkage-thresholding algorithm for linear inverse problems.SIAM Journal on Imaging Sciences, 2(1):183–202, 2009

    Amir Beck and Marc Teboulle. A fast iterative shrinkage-thresholding algorithm for linear inverse problems.SIAM Journal on Imaging Sciences, 2(1):183–202, 2009

    2009

  13. [13]

    Belloni, V

    A. Belloni, V . Chernozhukov, and L. Wang. Square-root lasso: pivotal recovery of sparse signals via conic programming.Biometrika, 98(4):791–806, 2011. 24

    2011

  14. [14]

    Bernardo and Adrian F

    José M. Bernardo and Adrian F. M. Smith.Bayesian Theory. Wiley, 1994

    1994

  15. [15]

    R., Linhart, J., Thirion, B., and Neuvial, P

    Alexandre Blain, Angel Reyero Lobo, Julia Linhart, Bertrand Thirion, and Pierre Neuvial. When knockoffs fail: diagnosing and fixing non-exchangeability of knockoffs.arXiv preprint arXiv:2407.06892, 2025

    work page arXiv 2025

  16. [16]

    SLOPE—Adaptive variable selection via convex optimization.The Annals of Applied Statistics, 9(3), 2015

    Małgorzata Bogdan, Ewout van den Berg, Chiara Sabatti, Weijie Su, and Emmanuel J Candès. SLOPE—Adaptive variable selection via convex optimization.The Annals of Applied Statistics, 9(3), 2015

    2015

  17. [17]

    Bondell and Brian J

    Howard D. Bondell and Brian J. Reich. Simultaneous Regression Shrinkage, Variable Selection, and Supervised Clustering of Predictors with OSCAR.Biometrics, 64(1):115–123, 2008

    2008

  18. [18]

    Group SLOPE – Adaptive Selection of Groups of Predictors.Journal of the American Statistical Association, 114(525):419–433, 2019

    Damian Brzyski, Alexej Gossmann, Weijie Su, and Małgorzata Bogdan. Group SLOPE – Adaptive Selection of Groups of Predictors.Journal of the American Statistical Association, 114(525):419–433, 2019

    2019

  19. [19]

    Gregor Buch, Andreas Schulz, Irene Schmidtmann, Konstantin Strauch, and Philipp S. Wild. Sparse Group Penalties for bi-level variable selection.Biometrical Journal, 66(4), 2024

    2024

  20. [20]

    Michael E Burczynski, Ron L Peterson, Natalie C Twine, Krystyna A Zuberek, Brendan J Brodeur, Lori Casciotti, Vasu Maganti, Padma S Reddy, Andrew Strahs, Fred Immermann, Walter Spinelli, Ulrich Schwertschlag, Anna M Slager, Monette M Cotreau, and Andrew J Dorner. Molecular Classification of Crohn’s Disease and Ulcerative Colitis Patients Using Transcripti...

    2006

  21. [21]

    BIV AS: A Scalable Bayesian Method for Bi-Level Variable Selection With Applications.Journal of Computational and Graphical Statistics, 29(1):40–52, 2020

    Mingxuan Cai, Mingwei Dai, Jingsi Ming, Heng Peng, Jin Liu, and Can Yang. BIV AS: A Scalable Bayesian Method for Bi-Level Variable Selection With Applications.Journal of Computational and Graphical Statistics, 29(1):40–52, 2020

    2020

  22. [22]

    Emmanuel Candès, Yingying Fan, Lucas Janson, and Jinchi Lv. Panning for Gold: ‘Model-X’ Knockoffs for High Dimensional Controlled Variable Selection.Journal of the Royal Statistical Society Series B: Statistical Methodology, 80(3):551–577, 2018

    2018

  23. [23]

    Two-stage subsampling variable selection for sparse high-dimensional generalized linear models.Statistical Methods in Medical Research, 34(7):1504–1521, 2025

    Marinela Capanu, Mihai Giurcanu, Colin B Begg, and Mithat Gönen. Two-stage subsampling variable selection for sparse high-dimensional generalized linear models.Statistical Methods in Medical Research, 34(7):1504–1521, 2025

    2025

  24. [24]

    Penalized regression, standard errors, and Bayesian lassos.Bayesian Analysis, 5(2):369–411, 2010

    George Casella, Malay Ghosh, Jeff Gill, and Minjung Kyung. Penalized regression, standard errors, and Bayesian lassos.Bayesian Analysis, 5(2):369–411, 2010

    2010

  25. [25]

    Bayesian Sparse Group Selection.Journal of Computational and Graphical Statistics, 25(3):665–683, 2016

    Ray-Bing Chen, Chi-Hsiang Chu, Shinsheng Yuan, and Ying Nian Wu. Bayesian Sparse Group Selection.Journal of Computational and Graphical Statistics, 25(3):665–683, 2016

    2016

  26. [26]

    Optimal two-step prediction in regression.Electronic Journal of Statistics, 11(1):2519–2546, 2017

    Didier Chételat, Johannes Lederer, and Joseph Salmon. Optimal two-step prediction in regression.Electronic Journal of Statistics, 11(1):2519–2546, 2017

    2017

  27. [27]

    The knockoff filter for FDR control in group-sparse and multitask regression

    Ran Dai and Rina Foygel Barber. The knockoff filter for FDR control in group-sparse and multitask regression. InProceedings of the 33rd International Conference on International Conference on Machine Learning, volume 48, pages 1851–1859. JMLR, 2016

    2016

  28. [28]

    Convergence of a stochastic approximation version of the EM algorithm.The Annals of Statistics, 27(1), 1999

    Bernard Delyon, Marc Lavielle, and Eric Moulines. Convergence of a stochastic approximation version of the EM algorithm.The Annals of Statistics, 27(1), 1999

    1999

  29. [29]

    L. H. Dicker. Variance estimation in high-dimensional linear models.Biometrika, 101(2): 269–284, 2014

    2014

  30. [30]

    Jianqing Fan, Shaojun Guo, and Ning Hao. Variance estimation using refitted cross-validation in ultrahigh dimensional regression.Journal of the Royal Statistical Society Series B: Statistical Methodology, 74(1):37–65, 2012

    2012

  31. [31]

    URLCRAN.R-project.org/package=sgs

    Fabio Feser.sgs: Sparse-group SLOPE: adaptive bi-level selection with FDR control, 2023. URLCRAN.R-project.org/package=sgs. R package version 0.3.9. 25

    2023

  32. [32]

    Sparse-group SLOPE: adaptive bi-level selection with FDR control.arXiv preprint arXiv:2305.09467, 2023

    Fabio Feser and Marina Evangelou. Sparse-group SLOPE: adaptive bi-level selection with FDR control.arXiv preprint arXiv:2305.09467, 2023

    work page arXiv 2023

  33. [33]

    Dual Feature Reduction for the Sparse-group Lasso and its Adaptive Variant

    Fabio Feser and Marina Evangelou. Dual Feature Reduction for the Sparse-group Lasso and its Adaptive Variant. InProceedings of the 42nd International Conference on Machine Learning, volume 267 ofProceedings of Machine Learning Research, pages 17068–17099. PMLR, 2025

    2025

  34. [34]

    Strong Screening Rules for Group-based SLOPE Models

    Fabio Feser and Marina Evangelou. Strong Screening Rules for Group-based SLOPE Models. InProceedings of The 28th International Conference on Artificial Intelligence and Statistics, volume 258 ofProceedings of Machine Learning Research, pages 352–360. PMLR, 2025

    2025

  35. [35]

    Mario A. T. Figueiredo and Robert D. Nowak. Sparse Estimation with Strongly Correlated Variables using Ordered Weightedℓ1 Regularization.arXiv preprint arXiv:1409.4005, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  36. [36]

    Foster and Edward I

    Dean P. Foster and Edward I. George. The Risk Inflation Criterion for Multiple Regression. The Annals of Statistics, 22(4):1947–1975, 1994

    1947

  37. [37]

    Regularization paths for generalized linear models via coordinate descent.Journal of Statistical Software, 33, 2010

    Jerome Friedman, Trevor Hastie, and Robert Tibshirani. Regularization paths for generalized linear models via coordinate descent.Journal of Statistical Software, 33, 2010

    2010

  38. [38]

    Selecting predictive biomarkers from genomic data.PLOS ONE, 17(6):1–21, 2022

    Florian Frommlet, Piotr Szulc, Franz König, and Malgorzata Bogdan. Selecting predictive biomarkers from genomic data.PLOS ONE, 17(6):1–21, 2022

    2022

  39. [39]

    Catch me if you can: signal localization with knockoff e-values.Journal of the Royal Statistical Society Series B: Statistical Methodology, page qkae042, 2024

    Paula Gablenz and Chiara Sabatti. Catch me if you can: signal localization with knockoff e-values.Journal of the Royal Statistical Society Series B: Statistical Methodology, page qkae042, 2024

    2024

  40. [40]

    Multilevel (hierarchical) modeling: What it can and cannot do.Technomet- rics, 48(3):432–435, 2006

    Andrew Gelman. Multilevel (hierarchical) modeling: What it can and cannot do.Technomet- rics, 48(3):432–435, 2006

    2006

  41. [41]

    George and Robert E

    Edward I. George and Robert E. McCulloch. Variable Selection Via Gibbs Sampling.Journal of the American Statistical Association, 88(423):881, 1993

    1993

  42. [42]

    George and Dean P

    EdwardI. George and Dean P. Foster. Calibration and empirical Bayes variable selection. Biometrika, 87(4):731–747, 2000

    2000

  43. [43]

    Asymptotic Properties of Bayes Risk of a General Class of Shrinkage Priors in Multiple Hypothesis Testing Under Sparsity.Bayesian Analysis, 11(3), 2016

    Prasenjit Ghosh, Xueying Tang, Malay Ghosh, and Arijit Chakrabarti. Asymptotic Properties of Bayes Risk of a General Class of Shrinkage Priors in Multiple Hypothesis Testing Under Sparsity.Bayesian Analysis, 11(3), 2016

    2016

  44. [44]

    Golub, Michael Heath, and Grace Wahba

    Gene H. Golub, Michael Heath, and Grace Wahba. Generalized cross-validation as a method for choosing a good ridge parameter.Technometrics, 21(2):215–223, 1979

    1979

  45. [45]

    Identification of Significant Genetic Variants via SLOPE, and its Extension to Group SLOPE

    Alexej Gossmann, Shaolong Cao, and Yu-Ping Wang. Identification of Significant Genetic Variants via SLOPE, and its Extension to Group SLOPE. InProceedings of the 6th ACM Conference on Bioinformatics, Computational Biology and Health Informatics, pages 232–240. Association for Computing Machinery, 2015

    2015

  46. [46]

    A Unified Framework for Pattern Recovery in Penalized and Thresholded Estimation and its Geometry.arXiv preprint arXiv:2307.10158, 2023

    Piotr Graczyk, Ulrike Schneider, Tomasz Skalski, and Patrick Tardivel. A Unified Framework for Pattern Recovery in Penalized and Thresholded Estimation and its Geometry.arXiv preprint arXiv:2307.10158, 2023

    work page arXiv 2023

  47. [47]

    Pinpointing Important Genetic Variants via A Feature-level Filter and Group Knockoffs.arXiv preprint arXiv:2408.12618, 2024

    Jiaqi Gu, Zhaomeng Chen, and Zihuai He. Pinpointing Important Genetic Variants via A Feature-level Filter and Group Knockoffs.arXiv preprint arXiv:2408.12618, 2024

    work page arXiv 2024

  48. [48]

    Unveiling low-dimensional patterns induced by convex non-differentiable regularizers.Annals of the Institute of Statistical Mathematics, 2025

    Ivan Hejný, Jonas Wallin, Małgorzata Bogdan, and Michał Kos. Unveiling low-dimensional patterns induced by convex non-differentiable regularizers.Annals of the Institute of Statistical Mathematics, 2025

    2025

  49. [49]

    McDonald

    Darren Homrighausen and Daniel J. McDonald. A study on tuning parameter selection for the high-dimensional lasso.Journal of Statistical Computation and Simulation, 88(15):2865–2892, 2018. 26

    2018

  50. [50]

    Model-free Knockoffs for SLOPE–Adaptive Variable Selection with Controlled False Discovery Rate

    Mahammad Humayoo and Xueqi Cheng. Model-free Knockoffs for SLOPE–Adaptive Variable Selection with Controlled False Discovery Rate. In2018 24th International Conference on Pattern Recognition (ICPR), pages 302–307. IEEE, 2018

    2018

  51. [51]

    Adaptive Bayesian SLOPE: Model Selection With Incomplete Data

    Wei Jiang, Małgorzata Bogdan, Julie Josse, Szymon Majewski, Bła ˙zej Miasojedow, and Veronika Roˇcková. Adaptive Bayesian SLOPE: Model Selection With Incomplete Data. Journal of Computational and Graphical Statistics, 31(1):113–137, 2022

    2022

  52. [52]

    Johnson and David Rossell

    Valen E. Johnson and David Rossell. Bayesian model selection in high-dimensional settings. Journal of the American Statistical Association, 107(498):649–660, 2012

    2012

  53. [53]

    Multilayer knockoff filter: Controlled variable selection at multiple resolutions.The Annals of Applied Statistics, 13(1), 2019

    Eugene Katsevich and Chiara Sabatti. Multilayer knockoff filter: Controlled variable selection at multiple resolutions.The Annals of Applied Statistics, 13(1), 2019

    2019

  54. [54]

    A flexible empirical Bayes approach to multiple linear regression, and connections with penalized regression.J

    Youngseok Kim, Wei Wang, Peter Carbonetto, and Matthew Stephens. A flexible empirical Bayes approach to multiple linear regression, and connections with penalized regression.J. Mach. Learn. Res., 25(1), 2024

    2024

  55. [55]

    Chemotherapy-induced dynamic gene expression changes in vivo are prognostic in ovarian cancer.British Journal of Cancer, 110:2975–2984, 2014

    A Koussounadis, S P Langdon, D J Harrison, and V A Smith. Chemotherapy-induced dynamic gene expression changes in vivo are prognostic in ovarian cancer.British Journal of Cancer, 110:2975–2984, 2014

    2014

  56. [56]

    Kremer, Sangkyun Lee, Małgorzata Bogdan, and Sandra Paterlini

    Philipp J. Kremer, Sangkyun Lee, Małgorzata Bogdan, and Sandra Paterlini. Sparse portfolio selection via the sortedℓ 1-norm.Journal of Banking & Finance, 110, 2020

    2020

  57. [57]

    The strong screening rule for SLOPE

    Johan Larsson, Małgorzata Bogdan, and Jonas Wallin. The strong screening rule for SLOPE. InAdvances in Neural Information Processing Systems, volume 33, pages 14592–14603. Curran Associates Inc., 2020

    2020

  58. [58]

    URL CRAN.R-project.org/ package=SLOPE

    Johan Larsson, Jonas Wallin, Malgorzata Bogdan, Ewout van den Berg, Chiara Sabatti, Emmanuel Candes, Evan Patterson, Weijie Su, Jakub Kała, Krystyna Grzesiak, and Michal Burdukiewicz.SLOPE: Sorted ℓ1 Penalized Estimation, 2022. URL CRAN.R-project.org/ package=SLOPE. R package version 0.5.0

    2022

  59. [59]

    Chapman and Hall/CRC, 2014

    Marc Lavielle.Mixed Effects Models for the Population Approach. Chapman and Hall/CRC, 2014

    2014

  60. [60]

    Don’t fall for tuning parameters: Tuning-free variable selection in high dimensions with the TREX.Proceedings of the AAAI Conference on Artificial Intelligence, 29(1), 2015

    Johannes Lederer and Christian Müller. Don’t fall for tuning parameters: Tuning-free variable selection in high dimensions with the TREX.Proceedings of the AAAI Conference on Artificial Intelligence, 29(1), 2015

    2015

  61. [61]

    A systematic review on model selection in high-dimensional regression.Journal of the Korean Statistical Society, 48(1):1–12, 2019

    Eun Ryung Lee, Jinwoo Cho, and Kyusang Yu. A systematic review on model selection in high-dimensional regression.Journal of the Korean Statistical Society, 48(1):1–12, 2019

    2019

  62. [62]

    Lee, Dennis L

    Jason D. Lee, Dennis L. Sun, Yuekai Sun, and Jonathan E. Taylor. Exact post-selection inference, with application to the lasso.The Annals of Statistics, 44(3):907–927, 2016

    2016

  63. [63]

    A note on the lasso and the related procedures in model selection.Statistica Sinica, 16(4):1273–1284, 2006

    Chenlei Leng, Yi Lin, and Grace Wahba. A note on the lasso and the related procedures in model selection.Statistica Sinica, 16(4):1273–1284, 2006

    2006

  64. [64]

    McDon- ald

    Xiaoxuan Liang, Aaron Cohen, Anibal Sólon Heinsfeld, Franco Pestilli, and Daniel J. McDon- ald. sparsegl: An R package for estimating sparse group lasso.Journal of Statistical Software, 110(6):1–23, 2024

    2024

  65. [65]

    Model-Free Feature Screening and FDR Control with Knockoff Features.Journal of the American Statistical Association, 117(537): 428–443, 2022

    Wanjun Liu, Yuan Ke, Jingyuan Liu, and Runze Li. Model-Free Feature Screening and FDR Control with Knockoff Features.Journal of the American Statistical Association, 117(537): 428–443, 2022

    2022

  66. [66]

    S. Lloyd. Least squares quantization in PCM.IEEE Transactions on Information Theory, 28 (2):129–137, 1982

    1982

  67. [67]

    Estimating the FDR of variable selection.arXiv preprint arXiv:2408.07231, 2024

    Yixiang Luo, William Fithian, and Lihua Lei. Estimating the FDR of variable selection.arXiv preprint arXiv:2408.07231, 2024. 27

    work page arXiv 2024

  68. [68]

    A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen.Cancer Cell, 5(6):607–616, 2004

    Xiao-Jun Ma, Zuncai Wang, Paula D Ryan, Steven J Isakoff, Anne Barmettler, Andrew Fuller, Beth Muir, Gayatry Mohapatra, Ranelle Salunga, J.Todd Tuggle, Yen Tran, Diem Tran, Ana Tassin, Paul Amon, Wilson Wang, Wei Wang, Edward Enright, Kimberly Stecker, Eden Estepa- Sabal, Barbara Smith, Jerry Younger, Ulysses Balis, James Michaelson, Atul Bhan, Karleen Ha...

    2004

  69. [69]

    Stability Selection.Journal of the Royal Statistical Society Series B: Statistical Methodology, 72(4):417–473, 2010

    Nicolai Meinshausen and Peter Bühlmann. Stability Selection.Journal of the Royal Statistical Society Series B: Statistical Methodology, 72(4):417–473, 2010

    2010

  70. [70]

    T. J. Mitchell and J. J. Beauchamp. Bayesian Variable Selection in Linear Regression.Journal of the American Statistical Association, 83(404):1023, 1988

    1988

  71. [71]

    The Cancer Genome Atlas Program, 1988

    National Cancer Institute. The Cancer Genome Atlas Program, 1988. URL cancer.gov/ about-nci/organization/ccg/research/structural-genomics/tcga

    1988

  72. [72]

    GAP Safe Screening Rules for Sparse-Group Lasso

    Eugene Ndiaye, Olivier Fercoq, Alexandre Gramfort, and Joseph Salmon. GAP Safe Screening Rules for Sparse-Group Lasso. InAdvances in Neural Information Processing Systems, volume 29. Curran Associates Inc., 2016

    2016

  73. [73]

    Feature screening and FDR control with knockoff features for ultrahigh- dimensional right-censored data.Computational Statistics & Data Analysis, 173:107504, 2022

    Yingli Pan. Feature screening and FDR control with knockoff features for ultrahigh- dimensional right-censored data.Computational Statistics & Data Analysis, 173:107504, 2022

    2022

  74. [74]

    Adaptive Three Operator Splitting

    Fabian Pedregosa and Gauthier Gidel. Adaptive Three Operator Splitting. InProceedings of the 35th International Conference on Machine Learning, volume 80 ofProceedings of Machine Learning Research, pages 4085–4094. PMLR, 2018

    2018

  75. [75]

    Splitting strategies for post-selection inference.Biometrika, 110(3):597–614, 2022

    D García Rasines and G A Young. Splitting strategies for post-selection inference.Biometrika, 110(3):597–614, 2022

    2022

  76. [76]

    A study of error variance estimation in lasso regression.Statistica Sinica, 26(1):35–67, 2016

    Stephen Reid, Robert Tibshirani, and Jerome Friedman. A study of error variance estimation in lasso regression.Statistica Sinica, 26(1):35–67, 2016

    2016

  77. [77]

    Derandomised knockoffs: leveraging e-values for false discovery rate control.Journal of the Royal Statistical Society Series B: Statistical Methodology, 86(1):122–154, 2024

    Zhimei Ren and Rina Foygel Barber. Derandomised knockoffs: leveraging e-values for false discovery rate control.Journal of the Royal Statistical Society Series B: Statistical Methodology, 86(1):122–154, 2024

    2024

  78. [78]

    Derandomizing Knockoffs.Journal of the American Statistical Association, 118(542):948–958, 2023

    Zhimei Ren, Yuting Wei, and Emmanuel Candès. Derandomizing Knockoffs.Journal of the American Statistical Association, 118(542):948–958, 2023

    2023

  79. [79]

    Brutlag, and Robert W

    Soo-Yon Rhee, Jonathan Taylor, Gauhar Wadhera, Asa Ben-Hur, Douglas L. Brutlag, and Robert W. Shafer. Genotypic predictors of human immunodeficiency virus type 1 drug resistance.Proceedings of the National Academy of Sciences, 103(46):17355–17360, 2006

    2006

  80. [80]

    Kremer, Sandra Paterlini, and Mał- gorzata Bogdan

    Riccardo Riccobello, Giovanni Bonaccolto, Philipp J. Kremer, Sandra Paterlini, and Mał- gorzata Bogdan. Sparse graphical modelling for minimum variance portfolios.SSRN Electronic Journal, 2023

    2023

Showing first 80 references.