Non-regular fractional factorial designs are widely used in screening experiments. Several analysis methods were proposed for identifying main effects and interactions in the past twenty years. The goal of this thesis is to evaluate the performance of two Bayesian and two frequentist approaches using some real data. The two Bayesian approaches are Bayes screening method and Bayesian hierarchical model selection method. The two frequentist approaches are traditional stepwise variable selection method and least angle regression method. The comparison of analysis methods is demonstrated with three examples of non-regular fractional factorial designs. Based on the models obtained from each analysis method, a slightly difference was observed for the Bayes screening method in one of our examples. However, both of the Bayesian approaches are flexible and efficient in identifying important effects. The traditional approach may be time-consuming to use for experiments with a large number of main factors and interactions. The least angle regression method is efficient but lack of flexibility because it only chooses models that satisfy the heredity principle.