Pith. sign in

REVIEW 3 major objections 7 minor 61 references

The paper claims that a thresholded ridge estimator recovers the support of the projected coefficient tensor in ultrahigh-dimensional tensor regression without sparsity, and that applying it to a 2,342x7,160x3 facial-shape tensor flags 2,39

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 04:45 UTC pith:772ZNYOJ

load-bearing objection Novel tensor-response screening method with a genuine gap: it selects on the projected coefficient B, not the SNP effect A, so the real-data loci claims and the stated theorem rate both need major revision. the 3 major comments →

arxiv 2607.22437 v1 pith:772ZNYOJ submitted 2026-07-24 stat.ME math.STstat.TH

A Consistent Feature Screening Approach for Tensor Responses with Applications to Genome-Wide Facial Shape Association

Shaofei Zhao , Zuofeng Shang , Seth M. Weinberg , Peter Claes , John R. Shaffer , Guifang Fu This is my paper
classification stat.ME math.STstat.TH MSC 62J0762F1262P10
keywords tensor responsefeature screeningridge regressiontrimmed estimatorselection consistencyultrahigh-dimensional datafacial shape GWASnon-sparse effects
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

TrimTenRidge is a feature-screening procedure for regression where the response is a multi-dimensional array (a tensor) and the predictors are ultrahigh-dimensional. The paper's central claim is that a ridge fit—least squares with an L2 penalty—followed by a threshold that zeroes small coefficients consistently separates strong from weak entries of the projected coefficient tensor, without any sparsity assumption. The proof delivers an explicit probability bound: with mild eigenvalue and boundedness conditions on the design and coefficient fibers, the thresholded support lies between the supports at slightly larger and slightly smaller thresholds with probability tending to 1, and the predictor dimension may grow at an exponential rate governed by the threshold. This matters for genome-wide facial-shape studies, where many SNPs have small nonzero effects and polygenic architecture makes sparsity unrealistic. Applied to a 2,342-by-7,160-by-3 tensor of facial shape from 2,342 people, the method flags 2,391 SNPs and associates each with facial regions.

Core claim

The discovery is that ridge regression, normally an estimation tool, can act as a consistent feature screener for tensor responses once its coefficients are thresholded. The paper works with B=<QQ^T,A>_{2,1}, the part of the coefficient tensor A that lies in the row space of the design matrix X; because A itself is not identifiable when p>>n, B is the identifiable target. With sub-Gaussian errors and conditions (C1)-(C2), choosing the ridge penalty h and threshold a_n in the rate window 0<alpha<(eta-tau-theta)/3 gives P(M_{B,a_nu_n} subset of M_{\hat B,a_n} subset of M_{B,a_n/u_n}) >= 1 - 4p exp{...}, so the trimmed support is sandwiched between strong and weak signal sets and no false posit

What carries the argument

The load-bearing machinery is the trimmed ridge estimator: \hat B = <(X^TX+hI)^{-1}X^T, Y>_{2,1}, with entries below a threshold a_n set to zero. The paper's key identity is the projection B = <QQ^T,A>_{2,1}, where X=PDQ^T is the SVD: it replaces the unidentifiable A with the identifiable part of the coefficient tensor lying in the row space of X. The computation avoids inverting a p x p matrix by using (X^TX+hI)^{-1}X^T = X^T(XX^T+hI)^{-1}, reducing the inverse to n x n. Theorem 2.2 then ties the threshold a_n and penalty h to the eigenvalue rate n^{-eta} and coefficient growth n^tau, producing a probability bound that permits ultrahigh-dimensional p.

Load-bearing premise

For the gene list, the load-bearing premise is that a SNP's importance can be read from the nonzero entries of B, the projection of the coefficient tensor onto the row space of X; with linkage disequilibrium this can inflate null SNPs and shrink causal ones, so support recovery for B does not automatically recover causal SNPs.

What would settle it

Use a design with two perfectly correlated SNP columns and set A nonzero only for the first SNP. TrimTenRidge assigns identical nonzero coefficients to both columns, so it flags a null SNP as strongly as the causal one, demonstrating that the recovered support is the correlation block's support, not the causal SNP's. For the theorem itself, simulate the Theorem 2.2 setup at the rate boundary with B entries just above a_n u_n: if the empirical inclusion probability fails to approach the stated lower bound over repeated samples, the bound is wrong.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Tensor-response screening no longer requires low-rank, CP, or Tucker sparsity structure, so small nonzero polygenic effects are not forced to zero.
  • Selected predictors come with a spatial signature: for each SNP, the nonzero coefficient entries identify which response regions—eyes, nose, lips, forehead, or whole face—are affected.
  • Jointly modeling all SNPs in a chromosome avoids the model misspecification of single-SNP GWAS caused by ignoring the polygenic background.
  • The facial-shape application yields a candidate set of 2,391 SNPs, including previously reported loci such as TASP1-related chin dimple and SRPK2-related eye/nose effects, plus new candidates for follow-up.
  • The consistency bound indicates that the maximum tolerable predictor dimension is log p = o(n^{2theta+2alpha}), giving an explicit growth limit for applied screening studies.

Where Pith is reading between the lines

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

  • Inference beyond the paper: the 2,391-SNP list should be read as linkage-disequilibrium-block membership, not individual causality. Because B projects A onto the row space of X, null SNPs in perfect LD with a causal SNP receive identical B entries, while a causal SNP whose genotype direction lies mostly outside the row space can have a small B entry.
  • Inference beyond the paper: a natural two-stage extension is to use TrimTenRidge as a screening stage and then run fine-mapping or conditional analysis inside each selected LD block to resolve actual causal variants; the paper does not perform this second stage.
  • Inference beyond the paper: the data-driven choice of a_n via prediction-error cross-validation selects the constant while the theory fixes the rate; an extension would derive the operating characteristics of that choice, since tiny nonzero effects are increasingly hard to distinguish as n grows and p grows faster.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The paper proposes TrimTenRidge, a ridge-regularized feature screening method for tensor responses with ultrahigh-dimensional matrix predictors. The core idea is to project the coefficient tensor A onto the row space of X, yielding B = ⟨QQ^T, A⟩_{2,1}, and then threshold the ridge estimate of B to select important predictor-response component combinations. The authors state that no sparsity assumption is needed, prove a selection-consistency theorem for the thresholded support of B, and apply the method to a facial-shape GWAS with a 2,342×7,160×3 tensor response and 6,322,729 SNPs, reporting both confirmatory and novel genetic loci.

Significance. If the proof gaps are repaired, the method would be a valuable screening tool for tensor-response regressions in ultrahigh dimensions. The computational trick reducing the p×p inverse to an n×n inverse is correct and practically important. The application to a large 3D facial-shape tensor is of broad interest, and the emphasis on polygenic, non-sparse effects is timely. However, the current manuscript has two load-bearing problems: (1) the theory recovers the support of the projected coefficient B, not the SNP-level coefficient A, so the real-data 'novel loci' claims are not supported; and (2) the main theorem's proof contains an algebraic error in the sub-Gaussian tail bound, inflating the claimed rate. These issues are fixable in principle but require a substantial revision.

major comments (3)
  1. [§2.2 and §4] The screening target is B = ⟨QQ^T, A⟩_{2,1}, the projection of the coefficient tensor A onto the row space of X. Models (1) and (2) are equivalent in conditional mean, so the transfer is sufficient for prediction, but not for variable selection. Theorem 2.2 and Corollary 2.3 only concern the thresholded support of B. The real-data section interprets nonzero entries of the estimator as 2,391 detected SNPs and calls them 'novel genetic loci.' Since B is a mixture of all entries of A through QQ^T, a null SNP in LD with a causal SNP can have a large B entry, and a causal SNP whose effect is partly orthogonal to R(X) can have a small or zero B entry. No identifiability condition (e.g., A ∈ R(X), or a suitable sparse-eigenvalue/irrepresentable condition) is stated or checked. The simulations generate B directly from model (2), so they validate recovery of B, not of A. The applied conclusion is
  2. [Appendix, Proof of Theorem 2.2] The tail bound is algebraically incorrect. With a_k^T = e_k^T (X^T X + hI)^{-1} X^T, a_k^T a_k = e_k^T (X^T X + hI)^{-1} X^T X (X^T X + hI)^{-1} e_k ≤ e_k^T (X^T X + hI)^{-1} e_k ≤ h^{-1}. Thus the linear combination a_k^T E_{:ij} has sub-Gaussian variance proxy σ²a_k^T a_k ≤ σ²h^{-1}. The proof instead uses |a_k^T a_k|² ≤ h^{-2} as the variance factor, which inflates the exponent by a factor of h. With h = M₂a_n^{-2}(log log n)n^θ and (u_n−1)a_n ~ a_n/log log n, the correct exponent is at most of order n^θ/(d₁d₂ log log n) (up to constants and union bounds), not n^{2θ+2α}. The stated rate log p = o(n^{2θ+2α}) is therefore unsupported. Please correct the proof and revise the theorem statement and its consequences accordingly.
  3. [Corollary 2.3] The corollary asserts exact support recovery P(Î = I*) → 1 under only mn = ∑_{I*}|B_{kij}| = √(d₁d₂)O(n^τ). No beta-min condition is imposed. Since a_n → 0, nonzero entries of B that are smaller than a_n u_n will be trimmed with probability not tending to zero (and asymptotically entirely if they are o(a_n)). A condition such as min_{I*}|B_{kij}| ≫ a_n is needed for Theorem 2.2 to imply exact recovery. As written, the corollary is not a consequence of Theorem 2.2.
minor comments (7)
  1. [§3] The text calls '1-specificity' the 'false negative rate'; this is actually the false positive rate (or 1-specificity). Please correct the terminology.
  2. [Supplementary Material] The text says 'in the following Tables 4-6' but the tables are labeled 2, 3, and 4. Please fix the numbering.
  3. [Proof of Lemma 2.1] The proof jumps from E∥bias∥ ≤ ... to E∥bias∥² = ... without justification; since the bias is deterministic given X, this should be stated. Also, the second term in part 2 appears to be h²λ₁^{-2}∥B∥² = O(h²n^{-2(η-τ)}), while the displayed rate is h²n^{-(1+η-2τ)}. The discrepancy should be resolved.
  4. [§4] The statement that rs6109993 and rs1479927 are 'ranked the highest by the TrimTenRidge approach' is not explained; no ranking procedure is defined in the methods section.
  5. [§5] The paper does not compare with any existing tensor-response method. Since Sun and Li (2017) and Li and Zhang (2017) address the same model class, a numerical comparison, even on the non-sparse settings, would strengthen the empirical contribution.
  6. [§2.2] The notation ⟨QQ^T, A⟩_{2,1} should be explicitly defined, since QQ^T is a matrix and A is a 3-way tensor; the definition is only implicit from the general inner-product notation.
  7. [Condition (C1)] The condition λ₁^{-1}=O_P(n^{-η}) is not verified for the real data. Since the rates in Theorem 2.2 depend on η, at least a heuristic check or discussion of its plausibility for the GWAS design matrix is needed.

Circularity Check

0 steps flagged

No significant circularity: the method estimates a projected coefficient B=QQ^T A; the real-data interpretation of B as SNP-level effects is an identifiability caveat, not a by-construction equivalence.

full rationale

The derivation chain is self-contained. Section 2.2 defines B = <QQ^T, A> and notes that XB = XA, so the ridge estimator targets the projected coefficient. This is an identifiability convention explicitly motivated by A not being identifiable; it is not a parameter fitted to the data, nor is the theorem's event used to define the estimator. Theorem 2.2 is a genuine bound: for h and a_n satisfying the stated rates, the thresholded ridge support lies between the large-|B| and small-|B| supports with probability at least 1 - 4p exp(-c1 n^{2theta+2alpha}/(2 d1^2 d2^2 sigma^2)). The proof uses bias/variance bounds and sub-Gaussian tail bounds; no equation in the theorem is identical to the estimator's construction. Corollary 2.3 asserts exact recovery of the support of B, not of A. Thus the simulations, which generate from model (2), coherently validate recovery of B. The real-data claim of 2,391 detected SNPs is a thresholded \hat B list; interpreting it as causal SNP associations requires an unstated assumption that support(B) approximates support(A) under LD, which is a correctness/external-validity risk rather than circularity. The only overlapping citation, Carlsen et al. (2016), supports background motivation (joint modeling and ridge choice) and is not load-bearing for the theorem; Shao et al. (2012) is external. No circular step is exhibited.

Axiom & Free-Parameter Ledger

2 free parameters · 5 axioms · 0 invented entities

The central result rests on the linear tensor model, the projection target B, sub-Gaussian noise, and two growth conditions (C1, C2). The threshold a_n and ridge h are chosen data-dependently in practice but treated as deterministic in theory. No new entities are posited.

free parameters (2)
  • threshold a_n = C n^{-alpha} = C and alpha chosen by CV over a candidate set; exact values not reported
    Central selection threshold; theorem treats it as deterministic with 0 < alpha < (eta - tau - theta)/3, but implementation tunes it from validation MSE.
  • ridge penalty h = h chosen by GCV; exact value not reported
    Used in the ridge estimator; theory sets h = M2 a_n^{-2} (log log n) n^theta, but the data-driven GCV choice is not analyzed.
axioms (5)
  • domain assumption Y = <X,A>_{2,1} + E with A the true coefficient tensor, and B = <QQ^T,A>_{2,1} the identifiable target.
    Section 2.2 defines B by projection to make the model identifiable; screening is for B, not A.
  • domain assumption E_{:ij}|X are i.i.d. sub-Gaussian with variance proxy sigma^2.
    Used in Lemma 2.1 and Theorem 2.2 for tail bounds; real genotype/phenotype noise may not be sub-Gaussian.
  • domain assumption Condition (C1): lambda_1^{-1} = O_P(n^{-eta}), eta <= 1.
    Requires the smallest positive eigenvalue of X^T X not too small; not verified for SNP matrices with LD.
  • domain assumption Condition (C2): ||B_{:ij}|| = O(n^tau), tau < eta.
    Bounded coefficient growth; avoids sparsity but limits coefficient magnitude growth.
  • standard math SVD and matrix inverse identities, sub-Gaussian concentration inequalities.
    Background for ridge estimator and tail bounds; standard results cited via Shao et al. 2012.

pith-pipeline@v1.3.0-alltime-deepseek · 18281 in / 26183 out tokens · 274874 ms · 2026-08-01T04:45:43.070166+00:00 · methodology

0 comments
read the original abstract

As data collecting technologies advance, data structures are getting more and more complex, from single vectors to multi-dimensional tensors. This article is motivated by a variable selection problem to detect important genes from an ultrahigh dimensional pool that are associated with human facial shape variations. We propose a data-driven trimmed feature screening method based on a tensor ridge regression model (TrimTenRidge) through setting thresholds on the tensor coefficients to perform a feature screening procedure. Unlike existing approaches, the TrimTenRidge does not require any sparse structures. In addition, it not only detects important predictors but also locates specific regions/components of the tensor response that are associated with each of the selected predictors. We prove the theoretical selection consistency and also assess its empirical performance through various simulation settings. The approach copes with ultra-high dimensional predictors and tensor responses simultaneously and contributes to the literature from theoretical, methodological, and five applicational aspects. We further apply the TrimTenRidge approach to genome-wide human facial shape data, from which the entire facial shapes form a $2,342\times 7,160\times 3$ tensor, and we successfully detect several novel genetic loci and also confirm some existing findings that are associated to facial shape.

Figures

Figures reproduced from arXiv: 2607.22437 by Guifang Fu, John R. Shaffer, Peter Claes, Seth M. Weinberg, Shaofei Zhao, Zuofeng Shang.

Figure 2
Figure 2. Figure 2: The false negative rates of TrimTenRidge approach for Simulation Setting 1 when assessing if it wrongly selects any of the p × d1 × d2 − 7 noise coefficients. As demonstrated in [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The true positive rates of TrimTenRidge approach for Simulation Setting 2 when assessing if it successfully selects all the 4 true coefficients. (  (  ( (   (   (  (  (  (  (            [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: The true positive rates of TrimTenRidge approach for Simulation Setting 3 when assessing if it successfully selects all the 15 true coefficients.. ( (  ( ( (  ( (  (   (  (     [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: The affected regions of five significant SNPs on chromosome 3 (rs6779419, rs7643249, etc.) [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: The affected region of SNP rs4980297 on chromosome 10 [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

61 extracted references

  1. [1]

    Nature , volume=

    Genome-wide association study of 107 phenotypes in Arabidopsis thaliana inbred lines , author=. Nature , volume=. 2010 , publisher=

    2010

  2. [2]

    The Annals of Statistics , volume=

    Estimation in high-dimensional linear models with deterministic design matrices , author=. The Annals of Statistics , volume=. 2012 , publisher=

    2012

  3. [3]

    SIAM Review , volume=

    Tensor decompositions and applications , author=. SIAM Review , volume=. 2009 , publisher=

    2009

  4. [4]

    Technometrics , volume=

    Generalized cross-validation as a method for choosing a good ridge parameter , author=. Technometrics , volume=. 1979 , publisher=

    1979

  5. [5]

    Nature Genetics , volume=

    Genome-wide mapping of global-to-local genetic effects on human facial shape , author=. Nature Genetics , volume=. 2018 , publisher=

    2018

  6. [6]

    Technical report KUL/ESAT/PSI/1401, KU Leuven, ESAT, Leuven, Belgium

    Development and comparison of non-rigid surface registration algorithms and extensions , author=. Technical report KUL/ESAT/PSI/1401, KU Leuven, ESAT, Leuven, Belgium. , year=

  7. [7]

    Modeling 3

    Claes, Peter and Liberton, Denise K and Daniels, Katleen and Rosana, Kerri Matthes and Quillen, Ellen E and Pearson, Laurel N and McEvoy, Brian and Bauchet, Marc and Zaidi, Arslan A and Yao, Wei and others , journal=. Modeling 3. 2014 , publisher=

    2014

  8. [8]

    Nature , volume=

    A global reference for human genetic variation , author=. Nature , volume=. 2015 , publisher=

    2015

  9. [9]

    Nature Methods , volume=

    Improved whole-chromosome phasing for disease and population genetic studies , author=. Nature Methods , volume=. 2013 , publisher=

    2013

  10. [10]

    PLoS Genetics , volume=

    A flexible and accurate genotype imputation method for the next generation of genome-wide association studies , author=. PLoS Genetics , volume=. 2009 , publisher=

    2009

  11. [11]

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

    Sure independence screening for ultrahigh dimensional feature space , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2008 , publisher=

    2008

  12. [12]

    Journal of the American statistical Association , volume=

    Variable selection via nonconcave penalized likelihood and its oracle properties , author=. Journal of the American statistical Association , volume=. 2001 , publisher=

    2001

  13. [13]

    The Journal of Machine Learning Research , volume=

    Ultrahigh dimensional feature selection: beyond the linear model , author=. The Journal of Machine Learning Research , volume=. 2009 , publisher=

    2009

  14. [14]

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

    Regression shrinkage and selection via the lasso , author=. Journal of the Royal Statistical Society: Series B (Methodological) , volume=. 1996 , publisher=

    1996

  15. [15]

    Journal of the royal statistical society: series B (statistical methodology) , volume=

    Regularization and variable selection via the elastic net , author=. Journal of the royal statistical society: series B (statistical methodology) , volume=. 2005 , publisher=

    2005

  16. [16]

    Journal of the American Statistical Association , volume=

    Model-free feature screening for ultrahigh-dimensional data , author=. Journal of the American Statistical Association , volume=. 2011 , publisher=

    2011

  17. [17]

    The Annals of Statistics , volume=

    A knockoff filter for high-dimensional selective inference , author=. The Annals of Statistics , volume=. 2019 , publisher=

    2019

  18. [18]

    Journal of Computational and Graphical Statistics , volume=

    Tensor-on-tensor regression , author=. Journal of Computational and Graphical Statistics , volume=. 2018 , publisher=

    2018

  19. [19]

    The Annals of Statistics , volume=

    Convex regularization for high-dimensional multiresponse tensor regression , author=. The Annals of Statistics , volume=. 2019 , publisher=

    2019

  20. [20]

    Thirty-first AAAI Conference on Artificial Intelligence , year=

    Multilinear regression for embedded feature selection with application to fmri analysis , author=. Thirty-first AAAI Conference on Artificial Intelligence , year=

  21. [21]

    BMC Bioinformatics , volume=

    Methods for the integration of multi-omics data: mathematical aspects , author=. BMC Bioinformatics , volume=. 2016 , publisher=

    2016

  22. [22]

    2010 IEEE International Conference on Image Processing , pages=

    Tensor completion for on-board compression of hyperspectral images , author=. 2010 IEEE International Conference on Image Processing , pages=. 2010 , organization=

    2010

  23. [23]

    IEEE Transactions on Signal Processing , volume=

    Tensor algebra and multidimensional harmonic retrieval in signal processing for MIMO radar , author=. IEEE Transactions on Signal Processing , volume=. 2010 , publisher=

    2010

  24. [24]

    Tensor decomposition for fast parsing with latent-variable

    Collins, Michael and Cohen, Shay , journal=. Tensor decomposition for fast parsing with latent-variable

  25. [25]

    BioMed research international , volume=

    Exploring the underlying genetics of craniofacial morphology through various sources of knowledge , author=. BioMed research international , volume=. 2016 , publisher=

    2016

  26. [26]

    Genome-wide association study of three-dimensional facial morphology identifies a variant in

    Paternoster, Lavinia and Zhurov, Alexei I and Toma, Arshed M and Kemp, John P and Pourcain, Beate St and Timpson, Nicholas J and McMahon, George and McArdle, Wendy and Ring, Susan M and Smith, George Davey and others , journal=. Genome-wide association study of three-dimensional facial morphology identifies a variant in. 2012 , publisher=

    2012

  27. [27]

    Nature Communications , volume=

    A genome-wide association scan implicates DCHS2, RUNX2, GLI3, PAX1 and EDAR in human facial variation , author=. Nature Communications , volume=. 2016 , publisher=

    2016

  28. [28]

    The American Journal of Human Genetics , volume=

    Disruption of contactin 4 (CNTN4) results in developmental delay and other features of 3p deletion syndrome , author=. The American Journal of Human Genetics , volume=. 2004 , publisher=

    2004

  29. [29]

    Journal of anatomy , volume=

    Reassessing the Dlx code: the genetic regulation of branchial arch skeletal pattern and development , author=. Journal of anatomy , volume=. 2005 , publisher=

    2005

  30. [30]

    Current topics in developmental biology , volume=

    Facial morphogenesis: physical and molecular interactions between the brain and the face , author=. Current topics in developmental biology , volume=. 2015 , publisher=

    2015

  31. [31]

    The Annals of Statistics , volume=

    The Dantzig selector: Statistical estimation when p is much larger than n , author=. The Annals of Statistics , volume=. 2007 , publisher=

    2007

  32. [32]

    The Annals of Statistics , volume=

    Lasso-type recovery of sparse representations for high-dimensional data , author=. The Annals of Statistics , volume=. 2009 , publisher=

    2009

  33. [33]

    The Annals of Statistics , volume=

    Simultaneous analysis of Lasso and Dantzig selector , author=. The Annals of Statistics , volume=. 2009 , publisher=

    2009

  34. [34]

    The Annals of Statistics , volume=

    The sparsity and bias of the lasso selection in high-dimensional linear regression , author=. The Annals of Statistics , volume=. 2008 , publisher=

    2008

  35. [35]

    Cell , volume=

    An expanded view of complex traits: from polygenic to omnigenic , author=. Cell , volume=. 2017 , publisher=

    2017

  36. [36]

    Weinberg, Seth M and Raffensperger, Zachary D and Kesterke, Matthew J and Heike, Carrie L and Cunningham, Michael L and Hecht, Jacqueline T and Kau, Chung How and Murray, Jeffrey C and Wehby, George L and Moreno, Lina M and others , journal=. The 3. 2016 , publisher=

    2016

  37. [37]

    Improved facial outcome assessment using a 3

    Claes, Peter and Walters, M and Clement, J , journal=. Improved facial outcome assessment using a 3. 2012 , publisher=

    2012

  38. [38]

    Gigascience , volume=

    Second-generation PLINK: rising to the challenge of larger and richer datasets , author=. Gigascience , volume=. 2015 , publisher=

    2015

  39. [39]

    Nature Genetics , volume=

    Detection and interpretation of shared genetic influences on 42 human traits , author=. Nature Genetics , volume=. 2016 , publisher=

    2016

  40. [40]

    Nature Genetics , volume=

    Insights into the genetic architecture of the human face , author=. Nature Genetics , volume=. 2021 , publisher=

    2021

  41. [41]

    Genetics in Medicine , volume=

    Elucidation of the phenotypic spectrum and genetic landscape in primary and secondary microcephaly , author=. Genetics in Medicine , volume=. 2019 , publisher=

    2019

  42. [42]

    New microdeletion and microduplication syndromes:

    Nevado, Juli. New microdeletion and microduplication syndromes:. Genetics and Molecular Biology , volume=. 2014 , publisher=

    2014

  43. [43]

    Genome scans of facial features in

    Liu, Chenxing and Lee, Myoung Keun and Naqvi, Sahin and Hoskens, Hanne and Liu, Dongjing and White, Julie D and Indencleef, Karlijne and Matthews, Harold and Eller, Ryan J and Li, Jiarui and others , journal=. Genome scans of facial features in. 2021 , publisher=

    2021

  44. [44]

    The Journal of Machine Learning Research , volume=

    STORE: sparse tensor response regression and neuroimaging analysis , author=. The Journal of Machine Learning Research , volume=. 2017 , publisher=

    2017

  45. [45]

    Journal of the American Statistical Association , volume=

    Parsimonious tensor response regression , author=. Journal of the American Statistical Association , volume=. 2017 , publisher=

    2017

  46. [46]

    Journal of the American Statistical Association , volume=

    Tensor regression with applications in neuroimaging data analysis , author=. Journal of the American Statistical Association , volume=. 2013 , publisher=

    2013

  47. [47]

    Statistics in Biosciences , volume=

    Tucker tensor regression and neuroimaging analysis , author=. Statistics in Biosciences , volume=. 2018 , publisher=

    2018

  48. [48]

    Manifold data analysis with applications to high-frequency 3

    Kang, Hyun Bin and Reimherr, Matthew and Shriver, Mark and Claes, Peter , journal=. Manifold data analysis with applications to high-frequency 3

  49. [49]

    2014 IEEE International Conference on Data Mining , pages=

    Tensor-based multi-view feature selection with applications to brain diseases , author=. 2014 IEEE International Conference on Data Mining , pages=. 2014 , organization=

    2014

  50. [50]

    Nature Biotechnology , volume=

    Context-aware dimensionality reduction deconvolutes gut microbial community dynamics , author=. Nature Biotechnology , volume=. 2021 , publisher=

    2021

  51. [51]

    Science Advances , volume=

    A GWAS in Latin Americans identifies novel face shape loci, implicating VPS13B and a Denisovan introgressed region in facial variation , author=. Science Advances , volume=. 2021 , publisher=

    2021

  52. [52]

    Hoskens, Hanne and Liu, Dongjing and Naqvi, Sahin and Lee, Myoung Keun and Eller, Ryan J and Indencleef, Karlijne and White, Julie D and Li, Jiarui and Larmuseau, Maarten HD and Hens, Greet and others , journal=. 3. 2021 , publisher=

    2021

  53. [53]

    Elife , volume=

    Novel genetic loci affecting facial shape variation in humans , author=. Elife , volume=. 2019 , publisher=

    2019

  54. [54]

    The Annals of statistics , volume=

    High dimensional classification using features annealed independence rules , author=. The Annals of statistics , volume=. 2008 , publisher=

    2008

  55. [55]

    Nature Genetics , volume=

    Shared heritability of human face and brain shape , author=. Nature Genetics , volume=. 2021 , publisher=

    2021

  56. [56]

    Genetics , volume=

    Exploiting linkage disequilibrium for ultrahigh-dimensional genome-wide data with an integrated statistical approach , author=. Genetics , volume=. 2016 , publisher=

    2016

  57. [57]

    2019 , publisher=

    Theory of ridge regression estimation with applications , author=. 2019 , publisher=

    2019

  58. [58]

    PLoS Genetics , volume=

    New entries in the lottery of facial GWAS discovery , author=. PLoS Genetics , volume=. 2016 , publisher=

    2016

  59. [59]

    A genome-wide association study identifies five loci influencing facial morphology in

    Liu, Fan and Van Der Lijn, Fedde and Schurmann, Claudia and Zhu, Gu and Chakravarty, M Mallar and Hysi, Pirro G and Wollstein, Andreas and Lao, Oscar and De Bruijne, Marleen and Ikram, M Arfan and others , journal =. A genome-wide association study identifies five loci influencing facial morphology in. 2012 , publisher=

    2012

  60. [60]

    Facial recognition from

    Sero, Dzemila and Zaidi, Arslan and Li, Jiarui and White, Julie D and Zarzar, Tom. Facial recognition from. Nature Communications , volume=. 2019 , publisher=

    2019

  61. [61]

    PLoS Genetics , volume=

    Genome-wide association study reveals multiple loci influencing normal human facial morphology , author=. PLoS Genetics , volume=. 2016 , publisher=

    2016