Data-Driven Variable Decomposition for Treatment Effect Estimation

One fundamental problem in causal inference is treatment effect estimation in observational studies when variables are confounded. Controlling for confounding effects is generally handled by propensity score. But it treats all observed variables as confounders and ignores the adjustment variables, which have no influence on treatment but are predictive of the outcome. Recently, it has been demonstrated that the adjustment variables are effective in reducing the variance of estimated effect. However, how to automatically separate the confounders and adjustment variables is still an open problem. In this paper, we firstly propose a Data-Driven Variable Decomposition (D2VD) algorithm, which can automatically separate confounders and adjustment variables, and simultaneously estimate treatment effect. Under standard assumptions, we theoretically prove that D2VD can unbiased estimate treatment effect with lower variance. Moreover, to address the nonlinear challenges, we propose a Nonlinear-D2VD (N-D2VD) algorithm. To validate their effectiveness, we conduct extensive experiments on both synthetic and real datasets. The results demonstrate that D2VD and N-D2VD algorithm can automatically separate the variables precisely, and estimate treatment effect more accurately and with lower variance than the state-of-the-art methods. We also demonstrated that the top-ranked features by our algorithm have the best prediction performance on an online advertising dataset.