报告摘要: In the estimation heterogeneous treatment effect, model uncertainty often exists because of the uncertainty which variable should be used or other reasons. To handle the model uncertainty, we propose a novel model averaging method for estimating the heterogeneous treatment effect by assembling the estimations from multiple additive candidate models with certain weights. The weights are obtained by minimizing J-fold cross-validation, in which nearest neighbor matching is used to impute the unobserved potential outcome. We show that the proposed method is asymptotically optimal in the sense of achieving the lowest possible squared loss and can put the weight one to the correctly specified models. Both simulation study and empirical example show the superiority of our proposed estimator over other competitive methods. |
