Selected publications

1. Debiased estimation:

   Demystifying Statistical Learning Based on Efficient Influence Functions Paper link

   Date: 2022

   Authors: Hines O, Dukes O, Diaz-Ordaz K & Vansteelandt S

   excerpt: This paper is about calculating efficient influence functions and using them to define debiased estimators that can use machine learning to estimate nuisance functionals.

2. Explainability and interpretability of causal effects:

   Variable importance measures for heterogeneous causal effects Paper link

   Authors: Hines O, Diaz-Ordaz K & Vansteelandt S

   excerpt: We present new nonparametric treatment effect variable importance measures (TE-VIMs). TE-VIMs can be viewed as nonparametric analogues to ANOVA but applied to a causal quantity. These may guide clinicians as to which patient characteristics are important to consider when making treatment decisions, and help researchers stratify patient populations for further study.

3. On locality of local explanation models Paper link

   Date: 2021

   Authors: Ghalebikesabi S, Ter-Minassian L, DiazOrdaz K, & Holmes C.

   excerpt: Shapley values provide model agnostic feature attributions for black-box models under a global population distribution. The use of a global population can be misleading when we are interested in local model behaviour. Hence we propose the use of neighbourhood reference distributions that improve the local interpretability.