arXiv Statistics ML
Robust inference for risk heterogeneity under group imbalance
Read the original on arXiv Statistics ML →The Flow has not summarised this story yet — read it at arXiv Statistics ML.
Related stories
arXiv AI
Oct 5
Distributionally Robust Survival Models under Subpopulation Shift and Outlier Contamination
The paper introduces a distributionally robust framework for survival analysis that simultaneously tackles latent subpopulation shift and outlier contamination. It employs an outer minimization to refine the nominal distribution by down-weighting contaminated samples and an inner maximization to target the most challenging subpopulation, directly handling non-decomposable survival losses such as the Cox partial log-likelihood. An alternating gradient-based algorithm, guided by KKT conditions, is developed, and experiments on simulated data and two benchmarks show improved worst-group performance and stable training under contamination.
arXiv Machine Learning
Oct 3
Beyond Demographic Balance: Multi-Metric and Intersectional Evaluation of Fairness in MIMIC-IV Mortality Prediction
The paper examines how fairness conclusions in ICU mortality prediction using MIMIC-IV depend on the choice of metrics and the granularity of demographic analysis. It compares predictive-utility and subgroup-error metrics across various fairness interventions and introduces a lightweight adaptation strategy that balances ethnicity, gender, and insurance representation without conditioning on mortality outcomes. The study finds that different interventions can be evaluated differently across accuracy, sensitivity, and false-positive rate, and that marginal demographic summaries may hide heterogeneous error patterns within intersectional subgroups.
arXiv Machine Learning
Jun 2
Scalable Counterfactual Risk Estimation for Rare Events in Longitudinal Data
arXiv:2606. 01539v1 Announce Type: cross Abstract: Estimating the causal effect of time-varying treatments on survival outcomes in large observational studies is computationally demanding, particularly when outcomes are rare.
arXiv AI
Jun 2
A Practical Upper Bound on Selection Bias Effects in Medical Prediction Models
arXiv:2606. 00563v1 Announce Type: cross Abstract: Selection bias is a common and often unavoidable aspect of real-world data that challenges the generalizability of machine learning models.
arXiv Machine Learning
Sep 4
Reliable Selection of Heterogeneous Treatment Effect Estimators
The paper introduces a method for selecting the best heterogeneous treatment effect (HTE) estimator from a set of candidates when the true treatment effect is unobserved. It frames estimator selection as a multiple testing problem and proposes a cross‑fitted, exponentially weighted test statistic that uses a two‑way sample splitting scheme to separate nuisance estimation from weight learning, ensuring stability for inference. The authors prove asymptotic familywise error rate control under mild conditions and demonstrate empirically that their procedure reduces false selections compared to common methods on ACIC 2016, IHDP, and Twins benchmarks.
arXiv Statistics ML
Sep 22
Bridging Local and Population Causal Effects: A Proximal Instrumental Variable Approach
arXiv:2609.22383v1 Announce Type: cross Abstract: Instrumental variable (IV) methods address treatment endogeneity, but with non-compliance and heterogeneous treatment effects a binary instrument gen...