Abstract:Estimating treatment effect plays an important role on decision making in many fields, such as social marketing, healthcare, and public policy. The key challenge on estimating treatment effect in the wild observational studies is to handle confounding bias induced by imbalance of the confounder distributions between treated and control units. Traditional methods remove confounding bias by re-weighting units with supposedly accurate propensity score estimation under the unconfoundedness assumption. Controlling high-dimensional variables may make the unconfoundedness assumption more plausible, but poses new challenge on accurate propensity score estimation. One strand of recent literature seeks to directly optimize weights to balance confounder distributions, bypassing propensity score estimation. But existing balancing methods fail to do selection and differentiation among the pool of a large number of potential confounders, leading to possible underperformance in many high dimensional settings. In this paper, we propose a data-driven Differentiated Confounder Balancing (DCB) algorithm to jointly select confounders, differentiate weights of confounders and balance confounder distributions for treatment effect estimation in the wild high dimensional settings. The synergistic learning algorithm we proposed is more capable of reducing the confounding bias in many observational studies. To validate the effectiveness of our DCB algorithm, we conduct extensive experiments on both synthetic and real datasets. The experimental results clearly demonstrate that our DCB algorithm outperforms the state-of-the-art methods. We further show that the top features ranked by our algorithm generate accurate prediction of online advertising effect.

Addressing Distribution Shift in RTB Markets via Exponential Tilting

Optimal Weighted Extreme Learning Machine for Imbalanced Learning with Differential Evolution [research Frontier]

ReTaSA: A Nonparametric Functional Estimation Approach for Addressing Continuous Target Shift

Training Unbiased Diffusion Models From Biased Dataset

Multiply Robust Estimation for Local Distribution Shifts with Multiple Domains

Decision-Focused Evaluation of Worst-Case Distribution Shift

Stable Heterogeneous Treatment Effect Estimation across Out-of-Distribution Populations

Chasing Fairness Under Distribution Shift: A Model Weight Perturbation Approach

Optimal Bidding Strategy without Exploration in Real-time Bidding

Adaptive Risk-Aware Bidding with Budget Constraint in Display Advertising

Rethinking Distribution Shifts: Empirical Analysis and Inductive Modeling for Tabular Data

Estimating Distributional Treatment Effects in Randomized Experiments: Machine Learning for Variance Reduction

Metalearners for Ranking Treatment Effects

Estimating Treatment Effect in the Wild via Differentiated Confounder Balancing

Deep Landscape Forecasting in Multi-Slot Real-Time Bidding

Improving uplift model evaluation on RCT data

Online and Scalable Model Selection with Multi-Armed Bandits

Advertisement Click-Through Rate Prediction Based on the Weighted-ELM and Adaboost Algorithm

Model-Agnostic Random Weighting for Out-of-Distribution Generalization

Stable Prediction with Model Misspecification and Agnostic Distribution Shift.

The Many Faces of Exponential Weights in Online Learning