mlsanityData to 14 Oct 2026

Feature Importance: A Closer Look at Shapley Values and LOCO

arXiv 2303.05981 ↗First posted 10 Mar 2023stat.ME stat.MLcited by 23

Not on the map. Why?

Abstract (arXiv)

There is much interest lately in explainability in statistics and machine learning. One aspect of explainability is to quantify the importance of various features (or covariates). Two popular methods for defining variable importance are LOCO (Leave Out COvariates) and Shapley Values. We take a look at the properties of these methods and their advantages and disadvantages. We are particularly interested in the effect of correlation between features which can obscure interpretability. Contrary to some claims, Shapley values do not eliminate feature correlation. We critique the game theoretic axioms for Shapley values and suggest some new axioms. We propose new, more statistically oriented axioms for feature importance and some measures that satisfy these axioms. However, correcting for correlation is a Faustian bargain: removing the effect of correlation creates other forms of bias. Ultimately, we recommend a slightly modified version of LOCO. We briefly consider how to modify Shapley values to better address feature correlation.

Key facts. Cited by 23 papers in the map (data through 14 Oct 2026). Builds on 16 of its 28 references that are in the map. First posted on arXiv 10 Mar 2023.

Builds on

16 of 28 references.

References · 16

  1. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  2. 2016Distribution-Free Predictive Inference For Regressioncited by 668
  3. 2016Model-Agnostic Interpretability of Machine Learningcited by 219
  4. 2020Explaining by Removing: A Unified Framework for Model Explanationcited by 218
  5. 2020Problems with Shapley-value-based explanations as feature importance measurescited by 200
  6. 2022Algorithms to estimate Shapley value feature attributionscited by 115
  7. 2014Controlling the false discovery rate via knockoffscited by 97
  8. 2019Unrestricted Permutation forces Extrapolation: Variable Importance Requires at least One More Model, or There Is No Free Variable Importancecited by 87
  9. 2020Efficient nonparametric statistical inference on population feature importance using Shapley valuescited by 61
  10. 2020Floodgate: inference for model-free variable importancecited by 16
  11. 2021Significance tests of feature relevance for a black-box learnercited by 15
  12. 2021Decorrelated Variable Importancecited by 13
4 more
  1. 2020Understanding Global Feature Contributions With Additive Importance Measurescited by 10
  2. 2021MDA for random forests: inconsistency, and a practical solution via the Sobol-MDAcited by 10
  3. 2021The HulC: Confidence Regions from Convex Hullscited by 6
  4. 2022Variable importance without impossible datacited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 2; 2024: 8; 2025: 8; 2026: 5.

23 papers.

Sort:

  1. 2023A decoder-only foundation model for time-series forecastingcited by 692
  2. 2024Statistical Significance of Feature Importance Rankingscited by 7
  3. 2023Explainable artificial intelligence approaches for brain-computer interfaces: a review and design spacecited by 5
  4. 2024Precise Model Benchmarking with Only a Few Observationscited by 5
  5. 2026ExplainerPFN: Towards tabular foundation models for model-free zero-shot feature importance estimationscited by 5
  6. 2024I Bet You Did Not Mean That: Testing Semantic Importance via Bettingcited by 3
  7. 2024Measuring Variable Importance in Heterogeneous Treatment Effects with Confidencecited by 3
  8. 2025Beyond Shapley Values: Cooperative Games for the Interpretation of Machine Learning Modelscited by 3
  9. 2025Doctor Rashomon and the UNIVERSE of Madness: Variable Importance with Unobserved Confounding and the Rashomon Effectcited by 3
  10. 2024Factor Importance Ranking and Selection using Total Indicescited by 2
  11. 2025Inference on Variable Importance for Treatment Effect Heterogeneity: Shapley Values and Beyondcited by 2
  12. 2024Energy-Based Model for Accurate Estimation of Shapley Values in Feature Attributioncited by 1
11 more
  1. 2025How to safely discard features based on aggregate SHAP valuescited by 1
  2. 2025Feature Attribution from First Principlescited by 1
  3. 2026Regional Explanations: Bridging Local and Global Variable Importancecited by 1
  4. 2024A new paradigm for global sensitivity analysiscited by 0
  5. 2024Targeted Learning for Variable Importancecited by 0
  6. 2025AICO: Feature Significance Tests for Supervised Learningcited by 0
  7. 2025Hierarchical Variable Importance with Statistical Control for Medical Data-Based Predictioncited by 0
  8. 2025Mathematical Theory of Collinearity Effects on Machine Learning Variable Importance Measurescited by 0
  9. 2026How important are the genes to explain the outcome - the asymmetric Shapley value as an honest importance metric for high-dimensional featurescited by 0
  10. 2026Attribution Upsampling should Redistribute, Not Interpolatecited by 0
  11. 2026Cluster LOCO: Feature Importance For Interpreting Clusterscited by 0