Estimating causal effect of a treatment on an outcome is often complicated. This is because the treatment effect may be deviated by the confounding variables. These variables affect treatment and outcome simultaneously, and the causal effect estimation thus depends upon these variables. Several methods have been proposed to reduce the attribute bias of confounding effect. In this article, we compare the traditional method of Propensity score through Stratification; and recent method of Propensity score through Bayesian in observational studies. Comparison is constructed on Mont Carlo simulation of the hypothetical binary treatment.