Measuring Variable Importance Using Proportional Inclusive Marginal Effects
Abstract
Shapley-based measures are widely used for measuring variable importance in statistics and machine learning. However, these measures have a serious issue--a nonactive variable can get a positive importance measure when it is correlated with an active variable. Herin et al. recently proposed a measure called proportional marginal effects (PME) that mitigates this spurious effect. However, PME is developed for deterministic models and has high computational cost. In this article, we propose a new measure called proportional inclusive marginal effects (PRIME) that overcomes these two limitations of PME. The key idea in PRIME is to use a novel model-free variable selection technique to identify the active variables and include only them in the computation of the importance measure. Analysis of test functions and real datasets show that PRIME can accurately and reliably quantify variable importance from noisy input-output data.
Related
- Just add noise: Debiasing tree-based variable importance in mixed data
- Marginal maximum likelihood estimation and asymptotic theory for latent variable models in high dimensions
- Sharp Instrumental Variable Bounds under The Marginal Sensitivity Model
- Measuring Negotiation Traits with Collaborative Choice Data: A Two-Stage Latent Variable Model
- Traveling Salesman Problem on Graphs with Proportional Independent Set
BibTeX
@misc{cho2026measuring,
title = {{Measuring Variable Importance Using Proportional Inclusive Marginal Effects}},
author = {Youngseo Cho and V. Roshan Joseph},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.17053},
eprint = {2610.17053},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.17053},
}
RIS
TY - UNPB
TI - Measuring Variable Importance Using Proportional Inclusive Marginal Effects
AU - Youngseo Cho
AU - V. Roshan Joseph
PY - 2026
DA - 2026/10/20
PB - arXiv
DO - 10.48550/arxiv.2610.17053
UR - https://arxiv.org/abs/2610.17053
AB - Shapley-based measures are widely used for measuring variable importance in statistics and machine learning. However, these measures have a serious issue--a nonactive variable can get a positive importance measure when it is correlated with an active variable. Herin et al. recently proposed a measure called proportional marginal effects (PME) that mitigates this spurious effect. However, PME is developed for deterministic models and has high computational cost. In this article, we propose a new measure called proportional inclusive marginal effects (PRIME) that overcomes these two limitations of PME. The key idea in PRIME is to use a novel model-free variable selection technique to identify the active variables and include only them in the computation of the importance measure. Analysis of test functions and real datasets show that PRIME can accurately and reliably quantify variable importance from noisy input-output data.
ER -