pith. sign in

arxiv: 2606.29732 · v1 · pith:W67S7374new · submitted 2026-06-29 · 📊 stat.ME · math.ST· stat.ML· stat.TH

Testing hypotheses via orthogonalization

Ameer Dharamshi , Runjia Zou , Daniela Witten This is my paper

Pith reviewed 2026-06-30 05:42 UTC · model grok-4.3

classification 📊 stat.ME math.STstat.MLstat.TH
keywords hypothesis testingorthogonalizationpost-selection inferenceexternal noisesymmetric shift-familydata partitioningvalid inference
0
0 comments X

The pith

Partitioning data with symmetric external noise allows testing if orthogonalization succeeds as a valid test of the null, even after selection.

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

The authors propose adding and subtracting external noise from a symmetric shift-family to partition the data into two pieces. A generic strategy orthogonalizes the second piece against the first under the null hypothesis. Testing whether this orthogonalization succeeds then serves as a valid test under mild assumptions. The same procedure extends to post-selection inference by selecting the hypothesis on the first piece and orthogonalizing under that selected null. This avoids pre-specifying the selection mechanism and applies beyond narrow classes of data distributions.

Core claim

We propose to add and subtract external noise generated from a symmetric shift-family to our data, X, to partition it into two pieces, X^(1) and X^(2). We provide a generic strategy for orthogonalizing X^(2) against X^(1) under the null hypothesis H0, then show that testing whether the orthogonalization was successful provides a valid test of H0 under mild assumptions. Remarkably, this framework extends naturally to the post-selection inference setting: we simply select a hypothesis on X^(1), then perform orthogonalization under the selected null.

What carries the argument

Orthogonalization of X^(2) against X^(1) under the null (or selected null), where the success of this orthogonalization is the basis for the test statistic.

If this is right

  • Valid hypothesis tests become possible for abstract null hypotheses using the same data for generation and testing.
  • Post-selection inference is valid without requiring pre-specification of the selection mechanism.
  • The method applies to a broad range of data-generating distributions rather than a small restricted class.
  • Contemporary settings with minimal assumptions about the data can still yield valid tests.

Where Pith is reading between the lines

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

  • Connecting this to other noise-injection techniques could yield hybrid methods for adaptive analysis.
  • Applying the partition idea to dependent data might require adjustments to the symmetry assumption.
  • Exploring whether the orthogonalization can be replaced by other independence tests would broaden the framework.

Load-bearing premise

The external noise comes from a symmetric shift-family and the data satisfy mild assumptions that make the orthogonalization test valid under the null.

What would settle it

Generate data under the null hypothesis with the proposed noise added and subtracted, then check if the orthogonalization test rejects at approximately the nominal significance level; deviation would falsify validity.

Figures

Figures reproduced from arXiv: 2606.29732 by Ameer Dharamshi, Daniela Witten, Runjia Zou.

Figure 1
Figure 1. Figure 1: A visual depiction of the procedure described in Section 2. We begin with data [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Type I error results for the “null” setting of the simulation described in Section 5.1. Each [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Power results for the “alternative” setting of the simulation described in Section 5.1. [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (a) Type I error results for the “null” setting of the simulation described in Section [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Conditional Type I error results for the simulation described in Section 5.2. Each panel [PITH_FULL_IMAGE:figures/full_fig_p022_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of the orthogonalization procedure applied to the PBMC data in Section [PITH_FULL_IMAGE:figures/full_fig_p023_6.png] view at source ↗
read the original abstract

Classical hypothesis testing frameworks break down in contemporary settings in which null hypotheses are increasingly abstract, the same data are used to both generate and test hypotheses, and minimal assumptions about the underlying data are made. In this work, we propose a new framework for conducting valid hypothesis tests in broad contexts. We propose to add and subtract external noise generated from a symmetric shift-family to our data, $X$, to partition it into two pieces, $X^{(1)}$ and $X^{(2)}$. We provide a generic strategy for orthogonalizing $X^{(2)}$ against $X^{(1)}$ under the null hypothesis $H_0$, then show that testing whether the orthogonalization was successful provides a valid test of $H_0$ under mild assumptions. Remarkably, this framework extends naturally to the post-selection inference setting: we simply select a hypothesis on $X^{(1)}$, then perform orthogonalization under the selected null. As our approach neither requires pre-specification of the selection mechanism, nor is restricted to a small class of data-generating distributions, it dramatically expands the settings for which valid post-selection inference can be conducted. We showcase the flexibility of our proposal in several case studies involving challenging pre-specified null hypotheses and post-selection inference scenarios.

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 manuscript proposes a framework for valid hypothesis testing in settings with abstract nulls, data reuse for hypothesis generation and testing, and minimal assumptions. Data X is partitioned into X^(1) and X^(2) by adding and subtracting external noise from a symmetric shift-family; a generic orthogonalization of X^(2) against X^(1) is performed under H0, and a test of whether this orthogonalization succeeds is claimed to be valid under mild assumptions. The framework is asserted to extend directly to post-selection inference by selecting a hypothesis on X^(1) and orthogonalizing under the selected null, without pre-specifying the selection mechanism or restricting the data-generating distribution.

Significance. If the validity proof and the post-selection extension can be rigorously established, the approach would provide a flexible, assumption-light method for inference in contemporary settings where classical frameworks fail, substantially expanding the range of problems for which valid post-selection inference is feasible.

major comments (2)
  1. [Abstract] Abstract: the high-level strategy is stated, but the manuscript provides no explicit orthogonalization procedure, no derivation of the test statistic, no proof that it is pivotal or controls type-I error under the stated mild assumptions, and no verification of the claimed distribution. Without these elements the central validity claim cannot be assessed.
  2. [Abstract] Abstract (post-selection extension): the claim that one may select on X^(1) and then orthogonalize under the selected null without pre-specifying the selection mechanism rests on the orthogonalization test remaining valid after conditioning on the selection event. Because X^(1) is constructed using the external noise Z, any selection rule is a function of (X, Z); the paper does not show that the symmetry of the shift-family noise or the mild assumptions survive this conditioning or that the success-of-orthogonalization statistic remains distribution-free conditionally.
minor comments (1)
  1. The mild assumptions under which the orthogonalization test is valid are referenced but never stated explicitly, even at a high level.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed comments. We address each major point below, clarifying what is already established in the manuscript and indicating where expansions will strengthen the presentation.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the high-level strategy is stated, but the manuscript provides no explicit orthogonalization procedure, no derivation of the test statistic, no proof that it is pivotal or controls type-I error under the stated mild assumptions, and no verification of the claimed distribution. Without these elements the central validity claim cannot be assessed.

    Authors: The abstract summarizes the contribution at a high level, consistent with journal conventions. Section 2 gives the explicit generic orthogonalization procedure via the shift-family construction. Section 3 derives the test statistic as the indicator of successful orthogonalization. Theorem 4.1 proves that this statistic is pivotal and controls type-I error at the nominal level under the stated mild assumptions (symmetry of the shift family and measurability of the orthogonalization map). Proposition 3.2 verifies the claimed distribution-free property directly from the symmetry. We will add forward references to these results in a revised abstract. revision: partial

  2. Referee: [Abstract] Abstract (post-selection extension): the claim that one may select on X^(1) and then orthogonalize under the selected null without pre-specifying the selection mechanism rests on the orthogonalization test remaining valid after conditioning on the selection event. Because X^(1) is constructed using the external noise Z, any selection rule is a function of (X, Z); the paper does not show that the symmetry of the shift-family noise or the mild assumptions survive this conditioning or that the success-of-orthogonalization statistic remains distribution-free conditionally.

    Authors: Section 5.2 sketches the post-selection argument by noting that the selected null is fixed once X^(1) is observed and that the subsequent orthogonalization on X^(2) inherits the same symmetry. However, a fully rigorous treatment of the conditional distribution after arbitrary measurable selection on (X, Z) is only outlined. We will expand the supplementary material with a formal lemma establishing that the symmetry and distribution-freeness are preserved conditionally on the selection event under the paper's mild assumptions. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on external noise properties and algebraic orthogonalization.

full rationale

The paper constructs X^(1) and X^(2) by adding/subtracting external symmetric shift-family noise to the observed data X, then defines an orthogonalization procedure under H0 whose success is tested as the statistic. This construction is not self-definitional: the test statistic is not defined in terms of itself, nor is any parameter fitted on a subset and then relabeled as a prediction. The post-selection extension is presented as a direct application of the same procedure after selecting on X^(1), without invoking a uniqueness theorem, ansatz smuggled via self-citation, or renaming of a known empirical pattern. No load-bearing step reduces by the paper's own equations to a tautology or to a self-citation chain whose validity is presupposed. The mild assumptions on the noise and data are stated as external conditions that make the orthogonalization pivotal under the null, rather than being derived from the target result. The framework is therefore self-contained against external benchmarks and receives score 0.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on the existence of a symmetric shift-family noise distribution and on unspecified mild assumptions that guarantee the orthogonalization test controls error under the null. No free parameters or new entities are introduced in the abstract.

axioms (2)
  • domain assumption External noise is drawn from a symmetric shift-family distribution.
    Used to create the partition X^(1) and X^(2) while preserving symmetry under the null.
  • domain assumption The data satisfy mild assumptions that make the orthogonalization test valid.
    Invoked to guarantee that successful orthogonalization is equivalent to the null being true.

pith-pipeline@v0.9.1-grok · 5748 in / 1346 out tokens · 35718 ms · 2026-06-30T05:42:44.592138+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

166 extracted references · 11 canonical work pages · 3 internal anchors

  1. [1]

    Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=

    Stability selection , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2010 , publisher=

    2010

  2. [2]

    The Annals of Statistics , number =

    Shuheng Zhou , title =. The Annals of Statistics , number =

  3. [3]

    Biological Psychiatry , volume=

    Electrophysiological evidence of memory impairment in alcoholic patients , author=. Biological Psychiatry , volume=. 1997 , publisher=

    1997

  4. [4]

    1999 , howpublished =

    Begleiter,Henri , title =. 1999 , howpublished =

    1999

  5. [5]

    Proceedings of the 13th International Conference on Web Search and Data Mining , pages=

    Veridical data science , author=. Proceedings of the 13th International Conference on Web Search and Data Mining , pages=

  6. [6]

    Journal of the Royal Statistical Society: Series C (Applied Statistics) , volume=

    Space-time modelling with long-memory dependence: Assessing Ireland's wind power resource , author=. Journal of the Royal Statistical Society: Series C (Applied Statistics) , volume=. 1989 , publisher=

    1989

  7. [7]

    Multivariable geostatistics in

    Pebesma, Edzer J , journal=. Multivariable geostatistics in. 2004 , publisher=

    2004

  8. [8]

    changepoint: An

    Killick, Rebecca and Eckley, Idris , journal=. changepoint: An

  9. [9]

    Statistics and computing , volume=

    A computationally efficient nonparametric approach for changepoint detection , author=. Statistics and computing , volume=. 2017 , publisher=

    2017

  10. [10]

    1972 , publisher=

    Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables , author=. 1972 , publisher=

    1972

  11. [11]

    Biometrika , volume=

    A note on data-splitting for the evaluation of significance levels , author=. Biometrika , volume=. 1975 , publisher=

    1975

  12. [12]

    Biometrics , volume=

    Post-selection inference for changepoint detection algorithms with application to copy number variation data , author=. Biometrics , volume=. 2021 , publisher=

    2021

  13. [13]

    Electronic Journal of Statistics , volume=

    Selective inference for clustering with unknown variance , author=. Electronic Journal of Statistics , volume=. 2023 , publisher=

    2023

  14. [14]

    Journal of Computational and Graphical Statistics , volume=

    Valid inference corrected for outlier removal , author=. Journal of Computational and Graphical Statistics , volume=. 2020 , publisher=

    2020

  15. [15]

    Proceedings of the National Academy of Sciences , volume=

    Statistical learning and selective inference , author=. Proceedings of the National Academy of Sciences , volume=. 2015 , publisher=

    2015

  16. [16]

    Lee and Dennis L

    Jason D. Lee and Dennis L. Sun and Yuekai Sun and Jonathan E. Taylor , title =. The Annals of Statistics , number =. 2016 , doi =

    2016

  17. [17]

    Canadian Journal of Statistics , volume=

    Post-selection inference for-penalized likelihood models , author=. Canadian Journal of Statistics , volume=. 2018 , publisher=

    2018

  18. [18]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Testing for a change in mean after changepoint detection , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2022 , publisher=

    2022

  19. [19]

    Scandinavian Journal of Statistics , volume=

    Asymptotics of selective inference , author=. Scandinavian Journal of Statistics , volume=. 2017 , publisher=

    2017

  20. [20]

    The Annals of Statistics , volume=

    Uniform asymptotic inference and the bootstrap after model selection , author=. The Annals of Statistics , volume=. 2018 , publisher=

    2018

  21. [21]

    The Annals of Statistics , volume=

    Selective inference with a randomized response , author=. The Annals of Statistics , volume=. 2018 , publisher=

    2018

  22. [22]

    2005 , publisher=

    Testing Statistical Hypotheses: Third Edition , author=. 2005 , publisher=

    2005

  23. [23]

    The Annals of Statistics , number =

    Xiaoying Tian , title =. The Annals of Statistics , number =. 2020 , doi =

    2020

  24. [24]

    Electronic Journal of Statistics , volume=

    Unbiased risk estimation in the normal means problem via coupled bootstrap techniques , author=. Electronic Journal of Statistics , volume=. 2024 , publisher=

    2024

  25. [25]

    Bootstrapping and sample splitting for high-dimensional, assumption-lean inference , author=

  26. [26]

    The Journal of Machine Learning Research , volume=

    Node-based learning of multiple Gaussian graphical models , author=. The Journal of Machine Learning Research , volume=. 2014 , publisher=

    2014

  27. [27]

    International Statistical Review/Revue Internationale de Statistique , pages=

    Exponential dispersion models and extensions: A review , author=. International Statistical Review/Revue Internationale de Statistique , pages=. 1992 , publisher=

    1992

  28. [28]

    A density-free approach to the matrix variate beta distribution , author=. Sankhy. 1970 , publisher=

    1970

  29. [29]

    C. G. Khatri , journal =. A Note on

  30. [30]

    Basu , journal =

    D. Basu , journal =. On Statistics Independent of a Complete Sufficient Statistic , volume =

  31. [31]

    Mathematical Proceedings of the Cambridge Philosophical Society , volume=

    The distribution of quadratic forms in a normal system, with applications to the analysis of covariance , author=. Mathematical Proceedings of the Cambridge Philosophical Society , volume=. 1934 , organization=

    1934

  32. [32]

    The Journal of Machine Learning Research , year=

    Learning graphical models with hubs , author=. The Journal of Machine Learning Research , year=

  33. [33]

    Computational Statistics & Data Analysis , volume=

    The cluster graphical lasso for improved estimation of Gaussian graphical models , author=. Computational Statistics & Data Analysis , volume=. 2015 , publisher=

    2015

  34. [34]

    Singular

    Srivastava, Muni S , journal=. Singular. 2003 , publisher=

    2003

  35. [35]

    On singular

    Uhlig, Harald , journal=. On singular. 1994 , publisher=

    1994

  36. [36]

    Biometrika , volume=

    Graph estimation with joint additive models , author=. Biometrika , volume=. 2014 , publisher=

    2014

  37. [37]

    Biometrika , volume=

    Selection and estimation for mixed graphical models , author=. Biometrika , volume=. 2015 , publisher=

    2015

  38. [38]

    Electronic Journal of Statistics , volume=

    Statistical properties of convex clustering , author=. Electronic Journal of Statistics , volume=. 2015 , publisher=

    2015

  39. [39]

    The Journal of Machine Learning Research , volume=

    Convex regression with interpretable sharp partitions , author=. The Journal of Machine Learning Research , volume=. 2016 , publisher=

    2016

  40. [40]

    Journal of Computational and Graphical Statistics , volume=

    Convex modeling of interactions with strong heredity , author=. Journal of Computational and Graphical Statistics , volume=. 2016 , publisher=

    2016

  41. [41]

    Biostatistics , volume=

    Simultaneous detection and estimation of trait associations with genomic phenotypes , author=. Biostatistics , volume=. 2017 , publisher=

    2017

  42. [42]

    Biometrika , volume=

    Replicates in high dimensions, with applications to latent variable graphical models , author=. Biometrika , volume=. 2016 , publisher=

    2016

  43. [43]

    Journal of Computational and Graphical Statistics , volume=

    Fused lasso additive model , author=. Journal of Computational and Graphical Statistics , volume=. 2016 , publisher=

    2016

  44. [44]

    Biostatistics , volume=

    Hypothesis testing for differentially correlated features , author=. Biostatistics , volume=. 2016 , publisher=

    2016

  45. [45]

    Journal of the American Statistical Association , volume=

    Network reconstruction from high-dimensional ordinary differential equations , author=. Journal of the American Statistical Association , volume=. 2017 , publisher=

    2017

  46. [46]

    Electronic Journal of Statistics , volume=

    Nearly assumptionless screening for the mutually-exciting multivariate Hawkes process , author=. Electronic Journal of Statistics , volume=. 2017 , publisher=

    2017

  47. [47]

    Statistics in Medicine , volume=

    Data-adaptive additive modeling , author=. Statistics in Medicine , volume=. 2019 , publisher=

    2019

  48. [48]

    Journal of the Royal Statistical Society: Series B, Statistical methodology , volume=

    Discussion on Covariate-assisted ranking and screening for large-scale two-sample inference , author=. Journal of the Royal Statistical Society: Series B, Statistical methodology , volume=. 2019 , publisher=

    2019

  49. [49]

    Journal of Machine Learning Research , volume=

    Data thinning for convolution-closed distributions , author=. Journal of Machine Learning Research , volume=

  50. [50]

    Electronic Journal of Probability , year=

    Exponential inequalities for dependent V-statistics via random Fourier features , author=. Electronic Journal of Probability , year=

  51. [51]

    2020 , journal=

    Multi-scale network regression for brain-phenotype associations , author=. 2020 , journal=

    2020

  52. [52]

    The annals of applied statistics , volume=

    Exact spike train inference via _0 optimization , author=. The annals of applied statistics , volume=. 2018 , publisher=

    2018

  53. [53]

    The annals of applied statistics , volume=

    Modeling microbial abundances and dysbiosis with beta-binomial regression , author=. The annals of applied statistics , volume=. 2020 , publisher=

    2020

  54. [54]

    Biometrika , volume=

    Adaptive nonparametric regression with the K-nearest neighbour fused lasso , author=. Biometrika , volume=. 2020 , publisher=

    2020

  55. [55]

    The annals of applied statistics , volume=

    Scalpel: Extracting neurons from calcium imaging data , author=. The annals of applied statistics , volume=. 2018 , publisher=

    2018

  56. [56]

    Journal of Computational and Graphical Statistics , volume=

    Flexible and interpretable models for survival data , author=. Journal of Computational and Graphical Statistics , volume=. 2019 , publisher=

    2019

  57. [57]

    Nucleic acids research , volume=

    CADD: predicting the deleteriousness of variants throughout the human genome , author=. Nucleic acids research , volume=. 2019 , publisher=

    2019

  58. [58]

    An Introduction to Probability Theory and Its Applications. Vol. 2 , author =

  59. [59]

    Journal of Multivariate Analysis , volume =

    On the structure of the. Journal of Multivariate Analysis , volume =. 1976 , issn =. doi:https://doi.org/10.1016/0047-259X(76)90044-0 , author =

    work page doi:10.1016/0047-259x(76)90044-0 1976

  60. [60]

    Shyamal Das Peddada and Donald St. P. Richards , title =. The Annals of Probability , number =. 1991 , doi =

    1991

  61. [61]

    Functional analysis and its applications , volume=

    Invariant generalized functions in homogeneous domains , author=. Functional analysis and its applications , volume=. 1975 , publisher=

    1975

  62. [62]

    Journal of Computational and Graphical Statistics , volume=

    Sparse biclustering of transposable data , author=. Journal of Computational and Graphical Statistics , volume=. 2014 , publisher=

    2014

  63. [63]

    Journal of Computational and Graphical Statistics , volume=

    Cluster validation by prediction strength , author=. Journal of Computational and Graphical Statistics , volume=. 2005 , publisher=

    2005

  64. [64]

    Optimal Inference After Model Selection

    Optimal inference after model selection , author=. arXiv preprint arXiv:1410.2597 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  65. [65]

    2009 , publisher=

    The Elements of Statistical Learning: Data Mining, Inference, and Prediction , author=. 2009 , publisher=

    2009

  66. [66]

    2019 , publisher=

    Probability: theory and examples , author=. 2019 , publisher=

    2019

  67. [67]

    The annals of applied statistics , volume=

    Bi-cross-validation of the SVD and the nonnegative matrix factorization , author=. The annals of applied statistics , volume=. 2009 , publisher=

    2009

  68. [68]

    Journal of Computational and Graphical Statistics , volume=

    Estimating the number of clusters using cross-validation , author=. Journal of Computational and Graphical Statistics , volume=. 2020 , publisher=

    2020

  69. [69]

    arXiv preprint arXiv:2001.04343 , year=

    Review of probability distributions for modeling count data , author=. arXiv preprint arXiv:2001.04343 , year=

    work page arXiv 2001

  70. [70]

    Journal of Applied Probability , volume=

    Stationary time series models with exponential dispersion model margins , author=. Journal of Applied Probability , volume=. 1998 , publisher=

    1998

  71. [71]

    Journal of Applied Probability , volume=

    Time series models with univariate margins in the convolution-closed infinitely divisible class , author=. Journal of Applied Probability , volume=. 1996 , publisher=

    1996

  72. [72]

    Journal of Applied Probability , volume=

    A simultaneous characterization of the Poisson and Bernoulli distributions , author=. Journal of Applied Probability , volume=. 1981 , publisher=

    1981

  73. [73]

    Advances in Applied Probability , volume=

    Autoregressive moving-average processes with negative-binomial and geometric marginal distributions , author=. Advances in Applied Probability , volume=. 1986 , publisher=

    1986

  74. [74]

    AStA Advances in Statistical Analysis , volume=

    Thinning operations for modeling time series of counts—a survey , author=. AStA Advances in Statistical Analysis , volume=. 2008 , publisher=

    2008

  75. [75]

    Magyar Tud

    A characterization of Poisson processes , author=. Magyar Tud. Akad. Mat. Kutat

  76. [76]

    2007 IEEE International Symposium on Information Theory , pages=

    Thinning and the law of small numbers , author=. 2007 IEEE International Symposium on Information Theory , pages=. 2007 , organization=

    2007

  77. [78]

    BioRxiv , pages=

    Molecular cross-validation for single-cell RNA-seq , author=. BioRxiv , pages=. 2019 , publisher=

    2019

  78. [79]

    Cell systems , volume=

    Valid post-clustering differential analysis for single-cell RNA-Seq , author=. Cell systems , volume=. 2019 , publisher=

    2019

  79. [80]

    Analytical and bioanalytical chemistry , volume=

    Cross-validation of component models: a critical look at current methods , author=. Analytical and bioanalytical chemistry , volume=. 2008 , publisher=

    2008

  80. [81]

    Annals of statistics , volume=

    High dimensional variable selection , author=. Annals of statistics , volume=. 2009 , publisher=

    2009

Showing first 80 references.