The propensity score (PS) is the probability of a subject receiving the treatment given the baseline covariates. People with the same propensity score tend to have the same distribution of covariates. Thus, propensity score related methods can be used to eliminate the systematic difference between treatment and control group so that improving the causal inferences in the observational study. In this project, a series of simulation studies are conducted to evaluate two widely used propensity score methods, matching and inverse probability of treatment weighting (IPTW), on their relative ability to estimate the treatment effect from non-randomized trials. One observes that the random forest based propensity score weighting can yield more promising treatment effect estimates compared with other PS methods. Besides that, simulated samples are also implemented to compare the performance of several matching methods on the balancing the covariates. It turns out that logistic regression based propensity score matching can reduce most of systematic differences between treatment and control group although it is not the top performer in the causal effect estimation. Finally, we illustrate the application of the propensity score methods discussed in the paper with an empirical example.