Additional file 1 of Estimation of average treatment effect based on a multi-index propensity score

Additional file 1: Fig. S1. The distribution of the estimated average treatment effect for kernel-based MiPS estimator and artificial neural network-based MiPS estimator in 1000 simulated data sets. The range of the y-axis is restricted from -1.4 to 0.6 given that the kernel-based MiPS estimator yields highly biased estimation under some model specifications. The dashed line denotes the true average treatment effect. Table S1. Estimation results for scenario with both continuous and discrete covariates under 50% treated based on 1000 replications. Table S2. Estimation results of multi-index propensity score estimator incorporating extra incorrect models under scenario with both continuous and discrete covariates. Table S3. Sensitivity analysis of ANN.MiPS estimator with different tuning parameters selection for ANN under scenario of all continuous covariates and 50% treated.