Balance Regularized Neural Network Models for Causal Effect Estimation
This work is significant for researchers and practitioners in healthcare and e-commerce who need to estimate causal effects from observational data, offering an incremental improvement to existing representation learning techniques.
This paper addresses the problem of estimating individual and average treatment effects from observational data by proposing a balance-regularized multi-head neural network architecture. The model reduces differences between treated and untreated distributions and encourages consistent control outcome predictions across groups, with an empirical study on bias-variance trade-offs and inference types.
Estimating individual and average treatment effects from observational data is an important problem in many domains such as healthcare and e-commerce. In this paper, we advocate balance regularization of multi-head neural network architectures. Our work is motivated by representation learning techniques to reduce differences between treated and untreated distributions that potentially arise due to confounding factors. We further regularize the model by encouraging it to predict control outcomes for individuals in the treatment group that are similar to control outcomes in the control group. We empirically study the bias-variance trade-off between different weightings of the regularizers, as well as between inductive and transductive inference.