ϕ-Table: A Statistical Explanation for Global SHAP

2025/12/08 by Kim, Dongseok, Choi, Hyoungsun, Rasool, Mohamed Jismy Aashik +1
Computer Science · #Bayesian Modeling and Causal Inference #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Methodology (stat.ME)

paper · doi:10.48550/arxiv.2512.07578

openalex publication_date 2025/12/08 · openalex created_date 2025/12/10 · openalex updated_date 2026/07/28

Abstract

Global SHAP explanations are typically presented as feature-importance rankings, which identify variables that matter to a black-box model but do not indicate whether their effects admit clear directional summaries, how uncertain those summaries are, or how faithfully they represent the fitted response. This paper proposes the ϕ-table, a SHAP-based statistical explanation table for tabular black-box regression models. The procedure selects features by SHAP importance and fits a standardized linear surrogate to the fitted model response f(X), reporting SHAP importance together with model-response coefficients, uncertainty summaries, surrogate fidelity, and bootstrap coefficient stability. The resulting coefficients are interpreted as projections of the fitted model response onto the SHAP-selected feature set. Across synthetic, semi-synthetic, and real-data experiments, the ϕ-table extends ranking-only SHAP into a statistical global explanation by exposing direction, uncertainty, fidelity, and stability as distinct components of fitted model behavior.

Citations

• Faithfulness Measurable Masked Language Models
• Explaining Predictive Uncertainty with Information Theoretic Shapley Values
• Evaluating the Robustness of Interpretability Methods through Explanation Invariance and Equivariance
• Robust Explanation Constraints for Neural Networks
• OpenXAI: Towards a Transparent Evaluation of Model Explanations
• Faith-Shap: The Faithful Shapley Interaction Index
• Framework for Evaluating Faithfulness of Local Explanations
• From Anecdotal Evidence to Quantitative Evaluation Methods: A Systematic Review on Evaluating Explainable AI
• Evaluating the Faithfulness of Importance Measures in NLP by Recursively Masking Allegedly Important Tokens and Retraining
• FastSHAP: Real-Time Shapley Value Estimation
• The effectiveness of feature attribution methods and its correlation with automatic evaluation scores
• Do Input Gradients Highlight Discriminative Features?
• Improving KernelSHAP: Practical Shapley Value Estimation via Linear Regression
• Explaining by Removing: A Unified Framework for Model Explanation
• Shapley Flow: A Graph-based Approach to Interpreting Model Predictions
• Reliable Post hoc Explanations: Modeling Uncertainty in Explainability
• Efficient nonparametric statistical inference on population feature importance using Shapley values
• Generalized and Scalable Optimal Sparse Decision Trees
• A general framework for inference on algorithm-agnostic variable importance
• CXPlain: Causal Explanations for Model Interpretation under Uncertainty
• Explanations can be manipulated and geometry is to blame
• Optimal Sparse Decision Trees
• The Shapley Taylor Interaction Index
• Sanity Checks for Saliency Maps
• A Benchmark for Interpretability Methods in Deep Neural Networks
• Towards Robust Interpretability with Self-Explaining Neural Networks
• Boolean Decision Rules via Column Generation
• Interpretation of Neural Networks is Fragile
• A Unified Approach to Interpreting Model Predictions
• What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?
• Axiomatic Attribution for Deep Networks
• Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
• Selective Inference for Group-Sparse Linear Models
• Scalable Bayesian Rule Lists
• "Why Should I Trust You?": Explaining the Predictions of Any Classifier
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
• Falling Rule Lists
• A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity

Related

• MinShap: A Shapley-Based Framework for Feature Redundancy
• Generalized SHAP: Generating multiple types of explanations in machine learning
• From Shapley Values to Generalized Additive Models and back
• A Unified Approach to Interpreting Model Predictions
• Explaining individual predictions when features are dependent: More accurate approximations to Shapley values