Paper: arXiv 2401.15139

Authors: Jasin Machkour, Daniel P. Palomar, Michael Muma

Abstract

In high-dimensional data analysis, such as financial index tracking or biomedical applications, it is crucial to select the few relevant variables while maintaining control over the false discovery rate (FDR). In these applications, strong dependencies often exist among the variables (e.g., stock returns), which can undermine the FDR control property of existing methods like the model-X knockoff method or the T-Rex selector. To address this issue, we have expanded the T-Rex framework to accommodate overlapping groups of highly correlated variables. This is achieved by integrating a nearest neighbors penalization mechanism into the framework, which provably controls the FDR at the user-defined target level. A real-world example of sparse index tracking demonstrates the proposed method’s ability to accurately track the S&P 500 index over the past 20 years based on a small number of stocks. An open-source implementation is provided within the R package TRexSelector on CRAN.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper presents sophisticated mathematical derivations including FDR control proofs and a novel nearest neighbors penalization mechanism (high math). It includes a real-world 20-year backtest on S&P 500 data, an open-source R package implementation, and mentions high-performance computing for empirical validation (high rigor).

Research Flowchart

  flowchart TD
  A["Research Goal: FDR-controlled sparse index tracking with dependent variables"] --> B["Propose Method: <br>Extended T-Rex with nearest neighbors penalization"]
  B --> C["Data: 20-year S&P 500 stock returns"]
  C --> D["Computation: <br>FDR-controlled variable selection via overlapping groups"]
  D --> E["Findings: <br>Accurate tracking with minimal stocks"]
  E --> F["Outcome: R package TRexSelector on CRAN"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a new FDR‑controlled sparse index‑tracking method that extends the dependency‑aware T‑Rex selector with a nearest‑neighbors penalization to handle overlapping groups of highly correlated stocks.

Source sentence
we propose a new FDR-controlling index tracking method by 1) extending the dependency-aware T-Rex selector [15] to account for strongly overlapping groups of highly correlated variables using a nearest neighbors penalization mechanism

Assumes 4

  • The method assumes a high‑dimensional setting with many variables (stocks). data
    Source sentence
    In high-dimensional data analysis, such as financial index tracking
  • Stock returns exhibit strongly overlapping groups of highly correlated variables. market
    Source sentence
    overlapping groups of highly correlated variables, which are characteristic for stock returns data
  • The theoretical FDR control holds as the number of knockoffs K tends to infinity. statistical
    Source sentence
    Theorem 1 (FDR control). Let Gr(j, ρthr) be as defined in (4) and K →∞.
  • The target FDR level α is specified by the investor and lies between 0 and 1. model
    Source sentence
    FDR(v, ρthr, T, L) ≤α, i.e., FDR control at the user-specified target level α ∈[0, 1]

Data

  • S&P 500 index tracking data (20‑year historical stock returns) market-prices unknown
    20 years · S&P 500 constituents
    Source sentence
    demonstrates the proposed method's ability to accurately track the S&P 500 index over the past 20 years based on a small number of stocks

Costs

Not modeled A trading or execution result is reported without costs.

Finds 2

  • The T‑Rex selector with nearest‑neighbors penalization (T‑Rex+DA+NN) achieves the best index‑tracking performance while using the fewest stocks in almost all quarters. performance
    Source sentence
    selector with the nearest neighbors penalization mechanism (T‑Rex+DA+NN) has the best index tracking performance, while requiring the smallest number of stocks in almost all quarters.
  • The proposed method tracks the S&P 500 over 20 years with higher accuracy and fewer stocks compared to state‑of‑the‑art methods, without needing sparsity‑parameter tuning. performance
    Source sentence
    has been successfully applied in tracking the S&P 500 index over the last 20 years with higher accuracy and fewer stocks compared to state-of-the-art methods. The proposed method does not require sparsity parameter tuning.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 5 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (106)

  1. Regression shrinkage and selection via the lasso (1996)
  2. Regularization and variable selection via the elastic net (2005)
  3. The adaptive lasso and its oracle properties (2006)
  4. Least angle regression (2004)
  5. Sparsity and smoothness via the fused lasso (2005)
  6. Pathwise coordinate optimization (2007)
  7. Controlling the false discovery rate: a practical and powerful approach to multiple testing (1995)
  8. The control of the false discovery rate in multiple testing under dependency (2001)
and 98 more
  1. Controlling the false discovery rate via knockoffs (2015)
  2. Panning for gold: ‘model- X ’ knockoffs for high dimensional controlled variable selection (2018)
  3. Probability with martingales (1991)
  4. Optimal stopping and free-boundary problems (2006)
  5. Biased sampling; the noncentral hypergeometric probability distribution (1963)
  6. Calculation methods for Wallenius’ noncentral hypergeometric distribution (2008)
  7. R: A Language and Environment for Statistical Computing (2019) — link
  8. knockoff: The Knockoff Filter for Controlled Variable Selection (2018) — link
  9. Regularization Paths for Generalized Linear Models via Coordinate Descent (2010)
  10. False discoveries occur early on the lasso path (2017)
  11. lars: Least Angle Regression, Lasso and Forward Stagewise (2013) — link
  12. A significance test for the lasso (2014)
  13. A knockoff filter for high-dimensional selective inference (2019)
  14. Robust inference with knockoffs (2018) — link
  15. The knockoff filter for FDR control in group-sparse and multitask regression (2016) — link
  16. Stability selection (2010)
  17. The sparsity and bias of the lasso selection in high-dimensional linear regression (2008)
  18. HIV-1 Protease and reverse-transcriptase mutations: correlations with antiretroviral therapy in subtype B isolates and implications for drug-resistance surveillance (2005)
  19. Genotypic predictors of human immunodeficiency virus type 1 drug resistance (2006)
  20. Deep knockoffs (2019)
  21. Exact post-selection inference, with application to the lasso (2016)
  22. The Bayesian lasso for genome-wide association studies (2011)
  23. A variable selection method for genome-wide association studies (2011)
  24. Ultrahigh dimensional feature selection: beyond the linear model (2009)
  25. A Markov blanket-based method for detecting causal SNP s in GWAS (2010)
  26. P-values for high-dimensional regression (2009)
  27. From disease association to risk assessment: an optimistic view from genome-wide association studies on type 1 diabetes (2009)
  28. How to interpret a genome-wide association study (2008)
  29. Replicability and prediction: lessons and challenges from GWAS (2018)
  30. Methods for handling multiple testing (2008)
  31. A selective overview of variable selection in high dimensional feature space (2010)
  32. Evaluation of the lasso and the elastic net in genome-wide association studies (2013)
  33. The post- GWAS era: From association to function (2018)
  34. Replicating genotype–phenotype associations (2007)
  35. Prioritizing GWAS results: a review of statistical methods and recommendations for their application (2010)
  36. 10 years of GWAS discovery: Biology, function, and translation (2017)
  37. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019 (2019)
  38. Examining the current standards for genetic discovery and replication in the era of mega-biobanks (2018)
  39. A power and prediction analysis for knockoffs with lasso statistics (2017) — link
  40. Confidence intervals and hypothesis testing for high-dimensional regression (2014)
  41. Near-ideal model selection by l1 minimization (2009)
  42. Asymptotics for lasso-type estimators (2000)
  43. On the “degrees of freedom” of the lasso (2007)
  44. Optimal inference after model selection (2014) — link
  45. SLOPE—adaptive variable selection via convex optimization (2015)
  46. A unified framework for high-dimensional analysis of $ M $-estimators with decomposable regularizers (2009)
  47. Controlling the rate of GWAS false discoveries (2017)
  48. Statistical learning and selective inference (2015)
  49. Hierarchical false discovery rate–controlling methodology (2008)
  50. Gene hunting with hidden Markov model knockoffs (2019)
  51. Unifying approach to selective inference with applications to cross-validation (2017) — link
  52. Information Theory and an Extension of the Maximum Likelihood Principle (1998) — link
  53. Estimating the dimension of a model (1978)
  54. The elements of statistical learning: data mining, inference, and prediction (2009)
  55. Selection of subsets of regression variables (1984)
  56. Variable selection with error control: another look at stability selection (2013)
  57. High dimensional variable selection (2009)
  58. Subset selection in regression (2002)
  59. Exact post-selection inference for sequential regression procedures (2016)
  60. Controlling variable selection by the addition of pseudovariables (2007)
  61. A note on data-splitting for the evaluation of significance levels (1975)
  62. Strong control, conservative point estimation and simultaneous conservative consistency of false discovery rates: a unified approach (2004)
  63. The positive false discovery rate: a Bayesian interpretation and the q-value (2003)
  64. U K B iobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age (2015)
  65. HAPGEN2 : simulation of multiple disease SNP s (2011)
  66. Integrating common and rare genetic variation in diverse human populations (2010)
  67. Detection of the number of signals using the Benjamini-Hochberg procedure (2007)
  68. A false discovery rate oriented approach to parallel sequential change detection problems (2020)
  69. Sparse activity detection for massive connectivity (2018)
  70. Direction of arrival estimation using co-prime arrays: A super resolution viewpoint (2014)
  71. Sparse distributed learning based on diffusion adaptation (2012)
  72. Sparse portfolios for high-dimensional financial index tracking (2017)
  73. Robust estimation in signal processing: A tutorial-style treatment of fundamental concepts (2012)
  74. Robust statistics for signal processing (2018)
  75. The outlier-corrected-data-adaptive Lasso: A new robust estimator for the independent contamination model (2017)
  76. A robust adaptive Lasso estimator for the independent contamination model (2020)
  77. Weakly convex regularized robust sparse recovery methods with theoretical guarantees (2019)
  78. The Terminating-Random Experiments Selector: Fast High-Dimensional Variable Selection with False Discovery Rate Control (2022) — link
  79. Derandomizing knockoffs (2021)
  80. tlars: The T-LARS Algorithm: E arly-Terminated Forward Variable Selection (2022) — link
  81. TRexSelector: T-Rex Selector: H igh-Dimensional Variable Selection & FDR Control (2022) — link
  82. Linkage disequilibrium in the human genome (2001)
  83. Sign-constrained least squares estimation for high-dimensional regression (2013)
  84. False Discovery Rate Control for Grouped Variable Selection in High-Dimensional Linear Models Using the T-Knock Filter (2022)
  85. False Discovery Rate ( FDR ) and Familywise Error Rate ( FER ) Rules for Model Selection in Signal Processing Applications (2022)
  86. … and the Cross-Section of Expected Returns (2015)
  87. Portfolio optimization and performance analysis (2007)
  88. Optimal benchmark tracking with small portfolios (2002)
  89. Index tracking with constrained portfolios (2007)
  90. Exact and heuristic approaches for the index tracking problem with UCITS constraints (2013)
  91. An efficient optimization approach for a cardinality-constrained index tracking problem (2016)
  92. High-Dimensional False Discovery Rate Control for Dependent Variables (2024) — link
  93. False Discovery Rate Control for Fast Screening of Large-Scale Genomics Biobanks (2023)
  94. sparseIndexTracking: Design of portfolio of stocks to track an index (2019) — link
  95. False Discovery Rate Control for G aussian Graphical Models via Neighborhood Screening (2024) — link
  96. Sparse PCA with False Discovery Rate Controlled Variable Selection (2024) — link
  97. Solving FDR -Controlled Sparse Regression Problems with Five Million Variables on a Laptop (2023)
  98. The Informed Elastic Net for Fast Grouped Variable Selection and FDR Control in Genomics Research (2023)

0 of 106 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.