Our purpose is to find which is the most reliable one among various oxygen abundance determination methods. We will test the validity of several different oxygen abundance determination methods using methods of modern statistics. These methods include Bayesian analysis and information scoring. We will analyze a sample of ∼ 6000 Hii galaxies from the Sloan Digital Sky Survey (SDSS) spectroscopic observations data release four. All methods that we used drew the same conclusion that the T e method is a more reliable oxygen abundance determination method than the Bayesian metallicity method under the existing telescope ability. The ratios of the likelihoods between the different kinds of methods tell us that the T e , P, and O3N2 methods are consistent with each other because the P and O3N2 methods are calibrated by T e method. The Bayesian and R 23 methods are consistent with each other because both are calibrated by a galaxy model. In either case, the N2 method is an unreliable method.