The purpose of this study is through Monte Carlo simulation to compare several propensity score methods in approximating factorial experimental design and identify best approaches in reducing bias and mean square error of parameter estimates of the main and interaction effects of two factors. Previous studies focused more on unbiased estimates of the effects of one factor, or the effects of one factor by the subgroups of another factor. The approaches for the unbiased estimates of the main and interaction effects of two factors in studies without full randomization were less examined. This study identifies appropriate propensity score methods to analyze multiple factors, in particular, the interaction effects. In sum, the simulation results from semi-experiment and non-experiment scenarios suggest three good propensity score applications in reducing bias and MSE of parameter estimates in analyzing two factors: (1) inverse of propensity score weighting based on one multinomial propensity score model, (2) subclassification and (3) factorial matching based on two binary propensity score models. Also note that the common support sample and covariate adjustment are preferred. (Contains 8 tables, 4 figures and 3 footnotes.)