Marginal and Conditional Importance Measures from Machine Learning Models and Their Relationship with Conditional Average Treatment Effect

Interpreting black-box machine learning models is challenging due to their strong dependence on data and inherently non-parametric nature. This paper reintroduces the concept of importance through "Marginal Variable Importance Metric" (MVIM), a model-agnostic measure of predictor importance based on the true conditional expectation function. MVIM evaluates predictors' influence on continuous or discrete outcomes. A permutation-based estimation approach, inspired by \citet{breiman2001random} and \citet{fisher2019all}, is proposed to estimate MVIM. MVIM estimator is biased when predictors are highly correlated, as black-box models struggle to extrapolate in low-probability regions. To address this, we investigated the bias-variance decomposition of MVIM to understand the source and pattern of the bias under high correlation. A Conditional Variable Importance Metric (CVIM), adapted from \citet{strobl2008conditional}, is introduced to reduce this bias. Both MVIM and CVIM exhibit a quadratic relationship with the conditional average treatment effect (CATE).

A Generalized Variable Importance Metric and Estimator for Black Box Machine Learning Models

The aim of this study is to define importance of predictors for black box machine learning methods, where the prediction function can be highly non-additive and cannot be represented by statistical parameters. In this paper we defined a ``Generalized Variable Importance Metric (GVIM)'' using the true conditional expectation function for a continuous or a binary response variable. We further showed that the defined GVIM can be represented as a function of the Conditional Average Treatment Effect (CATE) squared for multinomial and continuous predictors. Then we propose how the metric can be estimated using using any machine learning models. Finally we showed the properties of the estimator using multiple simulations.